IRON AND ALUMINIUM SPECIATION IN
SWEDISH FRESHWATERS
IMPLICATIONS FOR GEOCHEMICAL MODELLING
Carin Sjöstedt
June 2012
TRITA-LWR PHD 1065
ISSN 1650-8602
ISRN KTH/LWR/PHD 1065-SE
ISBN 978-91-7501-373-2
Carin Sjöstedt
TRITA LWR PHD 1065
© Carin Sjöstedt 2012
PhD thesis in Land and Water Resources Engineering
Royal Institute of Technology (KTH)
SE-100 44 STOCKHOLM, Sweden
5HIHUHQFH WR WKLV SXEOLFDWLRQ VKRXOG EH ZULWWHQ DV 6M|VWHGW & ´,URQ DQG
aluminium speciation in Swedish freshwaters. Implications for geochemical
PRGHOOLQJµ75,7$/:53+' 1065, 51 pp.
ii
Iron and aluminium speciation, Implications for geochemical modelling
´*ROGLVIRUWKHPLVWress - silver for the maid Copper for the craftsman cunning at his trade."
"Good!" said the Baron, sitting in his hall,
"But Iron - Cold Iron - is master of them all."
)URP´&ROG,URQµ5XG\DUG.LSOLQJ
iii
Carin Sjöstedt
TRITA LWR PHD 1065
iv
Iron and aluminium speciation, Implications for geochemical modelling
S UMMARY
Spårmetaller kan vara giftiga i höga koncentrationer. Totalkoncentrationerna av
spårmetaller i vatten har oftast inte något samband med biotillgängligheten;;
den senare styrs till största del av halten fria joner. Andelen fria joner av
totalhalten är därför en viktig faktor vid riskbedömningar av naturliga vatten.
Ett exempel där riskbedömningar av spårmetaller är viktiga är antropogent
försurade vatten. Varje år spenderar svenska staten 208 miljoner kronor på
kalkning av antropogent försurade ytvatten. Dock, så minskar den försurande
svaveldepositionen och på sikt bör kalkningen kunna upphöra. Ett hjälpmedel
för sådana framtidsscenarier och vid riskbedöming av ytvatten är geokemiska
modeller. Dessa grundar sig ofta på jämviktsuttryck mellan oorganiska joner,
en modell för bindningen till organiska humusämnen och en
adsorptionsmodell till (hydr)oxider. Två viktiga faktorer som styr andra
metallers speciering är metallerna järn och aluminium. De är vanliga i sura
vatten. De kan båda bilda hydroxider och oxider, som i sin tur kan adsorbera
andra spårmetaller vilket gör de senare mindre biotillgängliga. Järn och
aluminium binder också mycket bra till humusämnen. De kan därmed
konkurrera om bindningsplatser med andra metaller, vilket i sin tur gör dessa
metaller mer biotillgängliga. Trots mycket forskning är järn och aluminiums
kemi i naturliga vatten delvis okänd. Denna studie har som syfte att studera
järns och aluminiums former i fem svenska sjöar, för att bättre kunna simulera
aluminium och andra metallers speciering i ytvatten. Den förbättrade modellen
kan sedan användas för förutsägelser av vattenkemi efter avslutad kalkning i
Sverige.
Järn studerades kvalitativt genom att mäta järn(II) och genom µextended X-ray
absorption fine structureµEXAFS) spektroskopi i två näringsfattiga sjöar. Järn
och aluminium fraktionerades genom 0,45 µm-filtrering och 1 kDa-cross flow
(CF) ultrafiltrering, samt med dialys i ytterligare tre sjöar. Bly och koppars
adsorption till syntetisk ferrihydrit studerades med och utan förekomst av
fosfor i skakförsök och med EXAFS-spektroskopi. Resultaten visade att järn
delvis kunde finnas som järn(II) i syrerika ytvatten. Uppmätta partiklar (> 0,45
µm) i ett prov visade sig innehålla ferrihydrit. Vissa av kolloidproverna (<0,45
µm >1 kDa) innehöll även de delvis en järn(III)(hydr)oxid. Dock var kolloidalt
järn(III) framför allt komplexbundet till löst organiskt material (DOM) i
monomera komplex. Den monomera komplexbindningen var starkare än vad
som tidigare antagits. Komplexkonstanterna för järn(III)-bindning till
humusämnen i modellen Stockholm Humic Model (SHM) justerades därefter
och järns speciering simulerades. Den starkare komplexbindningen till DOM
medförde att järn(III) konkurrerade starkt med aluminium om bindningsplatser
enligt modellen. En modell konstruerades med SHM i programmet Visual
MINTEQ för att kunna simulera pH, oorganiskt monomert aluminium (Ali)
och andra metaller samtidigt. Först kalibrerades modellen för andelen aktivt
löst organiskt material i förhållande till löst organiskt kol (den så kallade
ADOM/DOC kvoten) genom att simulera pH via jonbalansen. Den kvot som
visade sig ge lägst fel var 1,65. Samma kvot och modelluppsättning användes
för att simulera halten Ali i flera stora dataset. Modellen gav ett tydligt samband
med uppmätta Ali-halter men med vissa modifieringar beroende på
analysmetod för Ali. Modellen kunde även simulera storleksfraktionering av
koppar och bly i två sjöar. Därefter användes modellen för att simulera
kalkavslut i de 3 043 kalkade sjöarna i Sverige. Modellresultaten visade att man
v
Carin Sjöstedt
TRITA LWR PHD 1065
bör kunna avsluta kalkningen i 30 % av sjöarna, framför allt i Norrland, och
även minska kalkningsverksamheten i andra delar av landet.
vi
Iron and aluminium speciation, Implications for geochemical modelling
A CKNOWLEDGEMENTS
First and foremost I would like to warmly acknowledge my main supervisor
Jon Petter Gustafsson, who has guided and encouraged me through these
years in an ever positive and constructive way. I feel very blessed with all the
help I have received throughout the years.
I also want to thank my excellent co-supervisors. Dan Berggren Kleja, for
cunning comments both on practical and theoretical things. Ingmar Persson,
for many (late) hours at Maxlab and for teaching me EXAFS theory. Hans
Borg, for guiding me both during my master thesis and PhD with great
knowledge about any trace metal and water chemistry issue. Martin
Hassellöv, for guidance in the project design.
My co-authors are also warmly acknowledged for their work with the papers.
Teresia Wällstedt, for being a brilliant supervisor for my master thesis and
introducing me to this particular world. Stephan Köhler, for always energetic
and humoristic discussions. Dean Hesterberg, for inviting me to stay at his
lab for three weeks and for all the help with EXAFS interpretations. Cecilia
Andrén, for giving discussions and for giving me new projects to work on all
the time. Charlotta Tiberg, for working hard these last weeks to be able to
include Paper V in my thesis (and for the rides to Uppsala).
To Ingrid Öborn at SLU, for employing me for the project in Vietnam and
for your always good advice. I am also thankful to the kind team in Vietnam
for all interesting discussions and hospitality.
I also owe a lot of credit to all the people who have helped me with the
laboratory work and performing analyses. At LWR I would like to
acknowledge Ann Fylkner, Monica Löwén and Bertil Nilsson for help with
all the instruments and analyses. The staffs at the Stockholm University, the
Dept. of Applied Environmental Science and the Dept. of Geological Sciences,
are kindly acknowledged for all the help and analyses performed, and
borrowing of the equipment. A special thanks to Ralf Dahlqvist for spending
many hours teaching me the ultrafiltration device. The staffs at SLU, Uppsala,
who have always been helpful. Especially Joris van Schaik, for teaching me
many new instruments. The staffs at MAX-lab, for much help at the beamline.
The staffs at Tyrestastiftelsen, for taking us to Årsjön in any kinds of weather.
The former and current PhD-students of the research group: Susanna, David,
Maja, and Satomi, for great feedback and much fun.
My colleagues at LWR, for all the good discussions and feedback over the
years and the nice working environment. Gunno Renman, for a constructive
review of my thesis. Aira, Britt, Joanne and Jerzy, for help with all practical
things related to being a PhD-student. To Mats and Zahra, my next door
neighbours, on which doors I can always knock for help (and coffee company).
There are many more which I have not mentioned by name, but to everyone
who has helped me through the years I am truly grateful!
Tack till vänner utanför forskarvärlden och till familjerna Sjöstedt och
Brynjer, för att ni alltid finns där. Sist men inte minst, min make Björn!
Härligt att vara förgiftad av/med dig!
The major part of this thesis has been performed at the Department of Land
and Water Resources Engineering, KTH Royal Institute of Technology.
Funding was mainly provided by the Swedish Research Council
(Vetenskapsrådet). Parts of the project were financed by the Swedish
vii
Carin Sjöstedt
TRITA LWR PHD 1065
Environmental Protection Agency. I was employed by SLU and funded by
SIDA for three months during my PhD-student period. The foundation A.W.
Bergstens donation is acknowledged for a travel stipend (H09-0055-AWB).
viii
Iron and aluminium speciation, Implications for geochemical modelling
TABLE
OF CONTENTS
Summary ............................................................................................................... v Acknowledgements ............................................................................................vii Table of contents ................................................................................................ ix List of appended papers and my contributions ............................................... xi Nomenclature and abbreviations .................................................................... xiii Abstract .................................................................................................................1 1. Introduction .................................................................................................1 1.1. Rationale for the thesis ..........................................................................1 1.1.1. 1.1.2. Toxicity and speciation of metals...........................................................................2 Objective and aims of this thesis ...........................................................................4 1.2. Review of previous work .......................................................................5 1.2.1. 1.2.2. 1.2.3. 1.2.4. 1.2.5. 2. Iron....................................................................................................................... 5 Aluminium ............................................................................................................9 Other trace metals ............................................................................................... 11 Natural organic matter ........................................................................................ 13 Mineral surfaces and surface complexation models .............................................. 15 Material and Methods ...............................................................................16 2.1. Site description and sampling ............................................................. 16 2.1.1. 2.1.2. Dalsland - (Paper I) ............................................................................................. 16 Tyresta ² (Paper III and Paper IV) ...................................................................... 16 2.2. Laboratory methods ............................................................................. 16 2.2.1. 2.2.2. 2.2.3. 2.2.4. Dialysis ² (Paper I) .............................................................................................. 16 In-line prefiltration ² (Paper III) ......................................................................... 17 Cross-flow ultrafiltration ² (Paper III) ................................................................. 17 Ferrozine method ² (Paper III) ........................................................................... 18 2.3. New method for concentrating of iron species ² Anion-exchange
column ............................................................................................................ 18 2.4. EXAFS spectroscopy ² (Paper III and Paper V) ............................... 18 2.4.1. 2.4.2. General ............................................................................................................... 18 EXAFS Data Analysis ......................................................................................... 19 2.5. Geochemical modelling ² (Paper I, II, III, IV, V) ............................ 20 2.5.1. 2.5.2. 2.5.3. 3. Data sets for modelling ² (Paper II and IV) ......................................................... 20 Modelling methods ............................................................................................. 21 Modelling scenarios for termination of liming ² (Paper IV) ................................. 23 Results........................................................................................................24 3.1. Iron: size fractionation and characterisation .....................................24 3.2. Aluminium: size fractionation and modelling in the five lakes ........26 3.3. The speciation of As and Mo in Dalsland ² (Paper I). ..................... 26 3.4. Copper(II) and lead(II) adsorption to ferrihydrite, as influenced by
phosphate ² (Paper V) ................................................................................... 27 3.5. A consistent geochemical model ² (Paper II and Paper IV) ............ 28 3.5.1. 3.5.2. 3.5.3. 3.5.4. Modelling pH ...................................................................................................... 28 Modelling inorganic monomeric aluminium ........................................................ 28 Modelling aluminium and iron particles ............................................................... 29 Modelling copper and lead fractionation .............................................................. 29 3.6. Predicted changes in lake chemistry at terminated liming ............. 30 4. Discussion .................................................................................................32 ix
Carin Sjöstedt
TRITA LWR PHD 1065
4.1. Iron speciation ..................................................................................... 32 4.2. The anion exchange column method ................................................. 34 4.3. Aluminium speciation .........................................................................35 4.3.1. Aluminium fractionation ..................................................................................... 37 4.4. Methods for studying trace metal speciation ....................................38 4.5. Geochemical equilibrium modelling ² is it possible to simulate pH
and metal speciation in lakes? ......................................................................39 4.6. The model as a tool for determining the extent of liming ² is it
possible to predict the future? ......................................................................41 4.7. Conclusions .......................................................................................... 42 References ...........................................................................................................43 Other references ............................................................................................. 51 x
Iron and aluminium speciation, Implications for geochemical modelling
L IST
OF APPENDED PAPE RS AND MY CONTRIBUTI ONS
Paper I
Sjöstedt, C., Wällstedt, T., Gustafsson, J.P., and Borg, H. (2009) Speciation of
aluminium, arsenic and molybdenum in excessively limed lakes. Science of the
Total Environment, 407, 5119-5127. Reprinted with permission from Elsevier.
I participated in planning, sampling and experimental work, performed the
geochemical modelling and the main part of data analysis and writing.
Paper II
Sjöstedt, C., Gustafsson, J.-P., and Köhler, S.J. (2010) Chemical equilibrium
modeling of organic acids, pH, aluminium, and iron in Swedish surface waters.
Environmental Science and Technology, 44 (22), 8587²8593. Reprinted with
permission. Copyright (2010) American Chemical Society.
I performed almost all modelling, data analysis and main part of writing.
Paper III
Sjöstedt, C., Persson, I., Hesterberg, D., Berggren Kleja, D., and Gustafsson,
J.P. Iron speciation in soft-water lakes and soils as determined by EXAFS
spectroscopy and geochemical modelling. Submitted to Geochimica et
Cosmochimica Acta 31 January 2012.
I performed most of the field and laboratory work, EXAFS and wavelet
interpretations, geochemical modelling and main part of writing.
Paper IV
Sjöstedt, C., Andrén, C, and Gustafsson, J.P. Modelling of pH and inorganic
aluminium after termination of liming in 3 000 Swedish lakes. Manuscript.
I performed all the modelling and the major part of data analysis and writing.
Paper V
Tiberg, C., Sjöstedt, C., Persson, I., and Gustafsson, J.P. Phosphate effects on
copper(II) and lead(II) sorption to ferrihydrite. Manuscript.
I was involved in EXAFS and wavelet interpretations and contributed to
writing.
xi
Carin Sjöstedt
TRITA LWR PHD 1065
xii
Iron and aluminium speciation, Implications for geochemical modelling
N OMENCLATURE
ADOM/DOC
Ali
CD-MUSIC
CF(F)
Colloids
DDL
DGT
DOC
DOM
EXAFS
FFF
ICP-OES
IR spectroscopy
NOM
OC
OM
Particles
PCV
SHM
TOC
XAS
AND ABBR EVIATIONS
Active dissolved organic matter to dissolved organic
carbon ratio
inorganic monomeric aluminium
The Charge distribution - multi site complexation model
cross-flow (filtration)
here defined as species being between 1 kDa and 0.45 µm
diffuse double layer
diffusive gradients in thin films
dissolved organic carbon
dissolved organic matter
extended X-ray absorption fine structure spectroscopy
field-flow fractionation
inductively coupled plasma- optical emission spectroscopy
infrared spectroscopy
natural organic matter
organic carbon
organic matter
Here defined as species larger than 0.4(5) µm
pyrochatechol violet
The Stockholm humic model
Total organic carbon
X-ray absorption spectroscopy
xiii
Carin Sjöstedt
TRITA LWR PHD 1065
xiv
Iron and aluminium speciation, Implications for geochemical modelling
A BSTRACT
Speciation governs transport and toxicity of trace metals and is important to
monitor in natural waters. Geochemical models that predict speciation are
valuable tools for monitoring. They can be used for risk assessments and
future scenarios such as termination of liming. However, there are often large
uncertainties concerning the speciation of iron and aluminium in the models,
due to the complicated chemistry of these metals. Both are important in
governing the speciation of other metals, due to (i) their capacity to form
minerals onto which metals can adsorb and (ii) their ability to compete for
binding sites to natural organic matter (NOM). Aluminium is also potentially
toxic and is therefore closely monitored in acidified freshwaters. In this study
different phases of iron in Swedish lakes were characterised. This required a
good method for preconcentrating the iron colloids. A new method was
developed in this thesis that uses an anion-exchange column to isolate the iron
colloids prior to characterisation with extended X-ray absorption fine structure
(EXAFS) spectroscopy. Iron was present as ferrihydrite in particles but was
also strongly monomerically complexed to NOM in two Swedish lakes. Based
on the results an internally consistent process-based geochemical equilibrium
model was presented for Swedish freshwaters. The model was validated for pH
(n = 9 400) and inorganic monomeric aluminium (Ali) (n = 3 400). The model
could simulate pH and Ali simultaneously, and be used for scenario modelling.
In this thesis, modelling scenarios for decreases and complete termination of
liming are presented for the 3 000 limed Swedish lakes. The results suggest that
liming can be terminated in 30 % of the Swedish lakes and decreased in many
other lakes.
Key words: Geochemical equilibrium modelling;; pH;; adsorption to
ferrihydrite;; metal-NOM complexation;; liming;; EXAFS spectroscopy
1. I NTRODUCTION
1.1. Rationale for the thesis
Trace metals in the environment are by definition present in small
concentrations, but they have a large impact. They can both be
essential to life but also toxic at quite low concentrations. They are
important to monitor due to their inability to be degraded. They
can however be immobilised by e.g. sedimentation and become less
available to biota. Their low concentrations make them a challenge
to study.
Metals have been used by Man since ancient times and they are an
important part of the human development. The problems that
metals cause are also known to have affected us for a long time.
The Roman Empire used large amounts of lead, e.g. for water
pipes. Even today it is possible to trace the Roman use of lead in
the sediments of Swedish waters. Hence, metal pollution is not
something new even though it has escalated during the industrial
ages. Some examples include the problems caused by acid rain
where the low pH of the soil and water cause leaching and elevated
concentration of potentially toxic metals such as aluminium;;
emissions of mercury and lead, arsenic poisoning due to natural
1
Carin Sjöstedt
TRITA LWR PHD 1065
high concentrations. Trace metals can be transported far from the
pollution source by air, and by ground and surface water.
Some of the aforementioned problems have caused significant
costs in human suffering, biological and economical degradation.
Trace metals are taken up by humans through drinking water, food
and air. Children and foetuses are especially vulnerable to lead and
mercury toxicity. Biologically there is a loss of biodiversity;; for
example, fish is especially vulnerable to acidity and cationic metals
such as aluminium. The Swedish state pays for instance 208
million SEK per year to counteract acidification with liming
(Swedish EPA, 2011).
For these reasons, monitoring of the potentially toxic trace metals
is of great importance. The good news about these toxicants is that
certain forms of them are not available to biota and are also not
transported far;; instead they are immobilised in the soil and in the
sediments. The conditions under which they are immobilised are
therefore of considerable interest. The focus of this thesis is to
improve the knowledge about how to ensure that the forms of
certain potentially toxic metals are in the non-toxic form for biota.
Other good news is that acidification is now on the retreat in
Europe and North America, due to lower emissions of sulphur
and nitrogen (Granier et al., 2011). Therefore it is of interest to
know if it is possible to terminate or decrease liming without
adverse effects on biota.
1.1.1. Toxicity and speciation of metals
For a long time it was assumed that the total concentration of the
toxic metal could be directly linked to toxicity. It was not until the
·V WKDW VRPH UHVHDUFKHUs suggested that it could be the free
ions that are the most important (Sunda and Guillard, 1976;; Morel,
1983). This gave rise to the free ion activity model (FIAM),
according to which the free ions are easily taken up by the
organisms. As a result they are the most important forms that
cause toxicity. On the other hand, if the ions are bound to natural
organic matter (NOM) or adsorbed to solids, they are no longer
free and therefore less bioavailable. The form of the metal, i.e. its
speciation, is accordingly of major interest. Later on this has been
further developed into the biotic ligand model (BLM) (di Toro et
al., 2001;; Niyogi and Wood, 2004), in which not only the free ion
activity of the toxic metal is important but also the competition of
other ions. Both the toxic and the non-toxic ions compete for
binding sites on the biotic ligands of the organism that takes up
the metal. Therefore not only the free ion concentration of the
toxic metal is important, but also the concentration and speciation
of other ions surrounding it. The pH value has an intricate
relationship to toxicity, since a lower pH generally means larger
solubility of metals and consequently more free ions, but at the
same time more hydrogen ions which can compete for binding
sites on the biotic ligand. This is complicated further when NOM
is present, which can bind the toxic metal as well as the competing
metal ions. The BLM approach for water quality criteria is now
2
Iron and aluminium speciation, Implications for geochemical modelling
currently used in regulations in the US for copper (US EPA, 2007),
and in the EU BLM:s can be used for deriving quality standards
(European Communities, 2011). It is therefore important that it is
valid. However, it is only valid if the relationships between all the
important factors are known. The FIAM and BLM are only valid
for situations where there is equilibrium in the surroundings of the
organism. The uptake of ions by biota is then a much slower
process than the diffusion in the environment (van Leeuwen et al.,
2005). For other situations diffusion may govern biological uptake
(van Leeuwen et al., 2005).
Much research has been focused upon determining the speciation
of metals and other important molecules in natural water (e.g.
Alfaro-de La Torre et al., 2000;; Gimpel et al., 2003;; Sigg et al.,
2006). There are many different speciation techniques available,
which use different principles for separating the different forms of
the metals. There are direct measurements of the free ion, e.g. by
potentiometry and ion-selective electrodes, but they are selective
and usually require high concentrations of the metal. There are
methods which separates molecules by size, since the smaller
fractions include the more potentially toxic free ions and dissolved
inorganic complexes. These methods include various forms of
ultrafiltration, dialysis, size-exclusion chromatography and fieldflow fractionation (FFF). Another separation technique is by
charge: e.g. cation- and anion-exchange columns. One challenge is
the high detection levels compared to natural concentrations. This
can be dealt with by preconcentration of the sample, e.g. by using
the Donnan membrane technique, by hollow fiber permeation
liquid membrane or by cross-flow ultrafiltration. When
determining the speciation, in-situ techniques are preferable due to
possible changes occurring during transport and at the laboratory.
However, there are few techniques available for in-situ speciation.
One in-situ device is diffusive gradients in thin films (DGT), but
this technique can cause problems with interpretation and is
difficult for low-ionic-strength soft-water (Alfaro-de la Torre et al.,
2000). Also, qualitative studies of the detailed structure of the
species by using large instruments are not possible in the field. All
methods have their advantages and limitations including problems
with contamination, changes during collection, storage and
treatment, and problems with interpretation. One thing they have
in common is that they are expensive, work-demanding and timeconsuming. They can therefore not be used for e.g. monitoring of
the 80 000 lakes in Sweden. They are also more or less
momentaneous, i.e. they reflect the speciation at one particular
point of time. To counteract these problems much work has been
devoted to producing geochemical models that can:
1. Estimate the free ion concentration and the other species in
water if the total concentration is known.
2. Predict what will happen in the future after changes of certain
chemical or physical parameters.
3
Carin Sjöstedt
TRITA LWR PHD 1065
Several attempts have been made to evaluate and compare
measured species of trace metals in surface water with simulated
ones using geochemical models (e.g. Gimpel et al., 2003;; Lyvén et
al., 2003;; Unsworth et al., 2006;; Vasyukova et al., 2010). The results
reported are variable. Some papers report that Cu and Pb are often
not correctly modelled (Gimpel et al., 2003;; Lyvén et al., 2003;;
Unsworth et al., 2006). Many of the conclusions focus on the role
of iron and aluminium which need to be better assessed to achieve
a ´goodµ model (Unsworth et al., 2006). Iron(III) and
aluminium(III) can form (hydr)oxides that can adsorb toxic metals
reducing their bioavailability. Both metals can also bind to NOM
and thereby compete with toxic metals for binding sites (Tipping et
al., 2002). This is especially important for soft-water environments
such as the Swedish lakes and rivers with a high Fe, Al and
dissolved organic carbon (DOC) content, which will affect the
other metals strongly. Without knowing the correct forms of iron
and aluminium, simulation of other metals becomes more difficult.
1.1.2. Objective and aims of this thesis
The objective of this thesis is to improve the understanding of the
speciation of iron and aluminium in surface water. This is
important to be able to more accurately assess the speciation of
other metals. My more specific aims are:
To determine if iron is present as a particulate, colloidal or truly
dissolved fraction in five Swedish lakes, to determine if it is
present as a mineral or an organic colloidal phase, and to
characterise the binding mode to dissolved organic matter (DOM)
in two Swedish lakes. To be able to characterise the binding mode
to DOM, a new method was needed for preconcentrating the iron
colloids (Paper I, Paper III)
To determine if aluminium is present as a particulate, colloidal or
truly dissolved fraction;; and if it is associated with a mineral or
with an organic colloidal/particulate phase in five Swedish lakes.
(Paper I, Paper III)
To determine the speciation of potentially toxic metals in
excessively limed lakes by comparing measured and modelled
fractions (Paper I)
To constrain and to test a consistent process-based geochemical
model concerning simulation of pH, aluminium, iron, and
adsorption of metals to ferrihydrite. The model should be both
internally consistent (the same parameters are used when
simulating an isolated one-component system as when using the
model for the whole system), and geochemically consistent with
results obtained earlier. (Paper II and V)
To use the model to predict what will happen with pH and
inorganic monomeric aluminium (Ali) at changes in lime doses in
the 3 043 limed Swedish lakes. (Paper IV)
4
Iron and aluminium speciation, Implications for geochemical modelling
1.2. Review of previous work
1.2.1. Iron
Iron is the second most abundant metal in the EDUWK·V FUXVW DQG
present in primary minerals such as silicates, oxides, and sulphides.
It is an essential element for many different kinds of organisms. It
is only toxic at very high concentrations to freshwater species (e.g.
Khan and Nugegoda, 2007 with references) compared to other
metals.
Most of the iron in surface waters originates from the catchment,
much less comes from atmospheric deposition (Lydersen et al.,
2002). Iron has a complicated chemistry in freshwaters and soils,
and can exist in the environment in two oxidation states, iron(II)
and iron(III). Iron(II) is the stable form in reducing environments
such as low-oxygen sediments, hypolimnetic lake water and
ground water. Iron(III) is the stable form in well-oxygenated
waters and soils. In oxygenated, circum-neutral pH water the
oxidation of iron(II) to iron(III) is fast, whereas in more acid water
the oxidation is slower (Davison, 1993). Organic matter and light
are known to be able to reduce iron(III) to iron(II) (Miles and
Brezonik, 1981). Few studies have measured iron(II) in oxygenated
surface water, even though it is not always negligible (Lofts et al.,
2008).
Iron(II) is generally present in the free hydrated ion form, since it
does not bind strongly to natural organic matter (van Dijk, 1971).
At significant concentrations and in the presence of other ions it
can precipitate in various mineral forms such as vivianite
[Fe3(PO4)2 · 8 H2O], siderite (FeCO3) or amorphous iron sulphide
(FeS) (Davison, 1993).
Iron(III) has a completely different chemistry. It has a very low
solubility;; it easily precipitates as different secondary oxides and
hydroxides. Examples of those are ferrihydrite, goethite >ơFeO(OH)], hematite >ơ-Fe2O3], lepidocrocite >ƣ-FeO(OH)], etc.
The iron(III) (hydr)oxides have some things in common. They can
easily adsorb other ions, both anions (due to positively charged
surfaces at environmental pH values) and cations that complex
specifically to the surfaces. This strong adsorption is due to a high
surface area and to a high amount of functional hydroxyl groups
with a pH-dependent charge. Another thing that distinguishes
iron(III) from iron(II) is that it readily complexes to natural
organic matter. These two features make iron(III) omnipresent in
particulate and colloidal form in freshwater.
Ferrihydrite is ubiquitous in the environment. It is nano-crystalline
and present in particles of sizes between 1 and 7 nm (Cismasu et al.
2011), which is smaller than the operationally used cut-off size for
particles, 0.45 µm. However, it can also aggregate into larger
particles. The exact structure is not known even for laboratoryproduced material. Recent research suggests the structure of a
single phase with a hexagonal space group P63mc and unit cell
dimensions of a = 5.95 Å and c = 9.06 Å (Michel et al., 2007)
(Fig. 1). The amount of tetrahedrally coordinated iron was later
5
Carin Sjöstedt
TRITA LWR PHD 1065
Fig. 1. Polyhedral representation of the hexagonal unit cell for
ferrihydrite. The bonded atoms (yellow) define a cubane-like
moiety that connects the basic structural motif of the model.
From Michel et al. (2007). Reprinted with permission from AAAS.
Readers may view, browse, and/or download material for temporary copying
purposes only, provided these uses are for noncommercial personal purposes.
Except as provided by law, this material may not be further reproduced,
distributed, transmitted, modified, adapted, performed, displayed, published,
or sold in whole or in part, without prior written permission from the publisher.
changed from 20 % to approximately 10 % (the remainder is
octahedral iron) due to higher cation vacancies and the
composition was adjusted to (Fe8.2O8.5(OH)7.4 · 3 H2O) (Michel et
al., 2010). The simpler formula of Fe(OH)3 is often used when
calculating the solubility product. The solubility product is also not
well known, but laboratory studies for the reaction
Fe(OH)3 (s) + 3 H+ ÅÆ Fe3+ + 3 H2O, *KS0
(1)
range between a log *KS0 of 2.5 to 5.0 (Tipping et al., 2002). Lower
values are determined for aged materials and higher for freshly
formed ferrihydrite (Tipping et al., 2002). The enthalpy of the
reaction has been determined to -100.4 kJ mol-1 by Liu and Millero
(1999).
However, the structure of the labotarory-synthesized ferrihydrite
can be different from the natural ferrihydrite, which in turn will
affect the adsorption capacities and the solubility. Other elements
present can substitute for Fe(III) or adsorb to the surface and
thereby hinder or affect polymerisation, e.g. Al, SiO4, or NOM
(Cismasu et al. 2011;; Liu and Hesterberg, 2011). Cismasu et al.
(2011) examined natural ferrihydrite from acid mine drainage
sediments. They found that crystallinity was reduced by adsorbed
6
Iron and aluminium speciation, Implications for geochemical modelling
Si and by a small amount of NOM. Silica was not believed to
chemically substitute into the structure due to the difference in
ionic radii between Si4+ and Fe3+. They also suspected Al to be
chemically substituted which could have caused additional
structural disorder.
On the contrary, Liu and Hesterberg (2011) precipitated Fe(III)
and Al in mixtures of 0, 20, 50, 75, 100 % Al of Al+Fe(III) at pH
7.5, and they did not found Al to substitute for Fe(III) in the
precipitates. Instead they interpreted their results for the 20 and
50 % Al samples to consist of core FeO6 polyhedra with AlO6
polyhedra integrated at their surfaces. The 75 % sample was
believed to consist of separate Al-hydroxide and Fe-hydroxide
phases, possibly aggregated together. Ferrihydrite also decreased in
structural order and domain size with increasing Al to Al+Fe(III)
ratio.
Iron (hydr)oxides can adsorb humic and fulvic acids (Tipping,
1981). Laboratory experiments show more humics to be adsorbed
at lower pH (Tipping, 1981). Schwertmann et al. (2005)
coprecipitated DOM and ferrihydrite which resulted in 96 mg C/g
ferrihydrite, as compared to results from adsorption experiments
where only between 3.3 and 24.2 mg C/g was adsorbed. They
interpreted this result as evidence for DOM being incorporated in
the structure. DOM reduced the crystallinity of ferrihydrite and
suppressed lepidocrocite formation.
When iron(III) and soil organic matter (OM) was mixed together
in the laboratory smaller ferrihydrite precipitates were formed than
with no OM present (Eusterhues et al. 2008). The amount of OM
that co-precipitated was 170 mg C per g ferrihydrite. However,
since sorption depends on the surface area, and smaller
precipitates were formed, the amount of C sorbed compared to
the surface area was similar as in other experiments (Eusterhues et
al., 2008). Neal et al. (2008) modelled the humic acid adsorption
with an empirical model, in which more humic substances were
adsorbed at lower pH to goethite. They then estimated the binding
of DOM to iron(III) oxyhydroxides in field river and lake water to
be around 5 %. For some sites there was a large temporal
variation.
Iron (hydr)oxide formation in three different waters have been
studied by Perret et al. (2000). Apparently the type of
macromolecular natural organic matter (aquagenic vs pedogenic;;
flexible vs compact) influences the morphology of the iron
particles formed in the oxic/anoxic zone.
Not just the structure and morphology but also the solubility
product of ferrihydrite can be different in the field compared to in
synthetic water. Lofts et al. (2008) calculated the solubility constant
for dialysed surface water samples and found that the slope of the
relationship between Fe3+ activity and pH described in (Eq. 1) was
lower compared to the stoichiometric value of -3. They suggested
possible explanations from the literature such as substitution of
7
Carin Sjöstedt
TRITA LWR PHD 1065
other anions than OH- in the solid phase or pH-dependent
variation in the activity of the solid.
The nature of iron(III) complexation to NOM is also not well
understood. Previous studies using extended X-ray absorption fine
structure (EXAFS) spectroscopy have indicated that iron can
complex both in monomeric (Karlsson et al., 2008;; van Schaik et
al., 2008;; Karlsson and Persson, 2010) and in dimeric/trimeric
forms (Rose et al., 1998;; Vilgé-Ritter et al., 1999;; Gustafsson et al.,
2007;; Mikutta and Kretzschmar, 2011) to organic matter. A
problem when using EXAFS spectroscopy for studying natural
samples is the high detection limit. To study the samples they
generally have to be preconcentrated, which could alter the
speciation. Complexation constants are also difficult to determine.
They have to be determined at low pH since iron(III) binds very
strongly at higher pH and to prevent formation of iron(III)
(hydr)oxides. A few studies exist;; Langford and Khan (1975)
studied iron(III) complexation to fulvic acid in a pH range from 1
to 2.5.
The strong complexation of Fe(III) makes it a competitor for
binding sites to organic matter for the metals Cu (Tipping et al.,
2002;; van Schaik et al., 2010), Zn, Cd, Ni, VO2+ (Gustafsson et al.,
2007) and probably also Pb (Tipping, 2005). This would increase
the presence of these metals in their dissolved inorganic form
which makes them more bioavailable.
Previous studies of iron and its correlations with DOC and with
other trace metals in surface water give a very mixed result
depending on origin, type of surface water, season, discharge, etc.
Iron was positively correlated to DOC in cloud water, rainfall,
throughfall and stemflow, streams, but not to groundwater in the
UK (Neal et al., 2008). Increasing iron concentrations in streams
and lake have been notified in the UK together with DOC
increases, but this was claimed not to be due to complexation of
Fe(III) to DOM. Instead they explained it by colloidal stabilization
of iron (hydr)oxide particles by organic matter, which would cause
them to be more easily transported. Some studies of size
fractionation do not show a correlation between iron and DOC,
instead iron colloids and DOC colloids were seen as two different
pools in circum-neutral boreal Russian rivers, streams, swamps and
lakes (Pokrovsky and Schott, 2002;; Vasyukova et al., 2010).
Aluminium and lead were associated with iron, probably due to coprecipitation according to the authors.
Lyvén et al. (2003) studied a pH 6.5 creek water with a DOC
concentration of 4 mg L-1 and an iron concentration (< 0.45 µm)
of 884 µg L-1. Here iron and DOC were present in two size
fractions, where the iron colloids (~ 4 nm) were larger than the
DOC colloids (~1.2 nm). An intermediate fraction also existed.
To sum up, iron influences the transport and toxicity of several
metals. This is caused either by competition for binding sites to
organic matter or by the formation of (hydr)oxides to which
8
Iron and aluminium speciation, Implications for geochemical modelling
metals can bind. It is however a complex relationship since iron
(hydr)oxides also adsorb organic matter.
This study will investigate in what forms iron resides in Swedish
lakes, by use of size separation, Fe(II) determination and
characterisation by EXAFS spectroscopy. A new method was
developed to preconcentrate the colloids for EXAFS spectroscopy
using an anion-exchange column. Also, adsorption to ferrihydrite
is studied in detail for Cu and Pb by batch experiments and by
EXAFS spectroscopy. The results are used to improve
geochemical modelling of other metals such as aluminium, copper
and lead.
1.2.2. Aluminium
Aluminium is the third most abundant element and the most
FRPPRQ PHWDO LQ WKH (DUWK·V FUust. It is present in granites and
gneisses. So far it has not been proved that aluminium is an
essential element to any organism. Instead it is toxic at large
concentrations, especially to fish, but also to other aquatic
organisms e.g. invertebrates and plankton (for a review, see e.g.
Gensemer and Playle, 1999).
The major potentially toxic forms of aluminium are the inorganic
cations, especially the free Al3+ ion. Two types of toxic
mechanisms are suggested for fish;; ionoregulatory problems at
lower pH due to Al3+ displacing Ca2+ at tight junctions or
interfering with Na, K-ATPase (Gensemer and Playle, 1999). This
is similar to the potential toxicity of the H+ ion and other metal
cations. The other mechanism is precipitation and/or
polymerisation of Al at the gills causing respiratory problems. This
is most severe at pH around 6 (Gensemer and Playle, 1999). High
Ca2+, H+ and DOM counteract the problem with potential Al
toxicity for fish in accordance with BLM theory.
Atmospheric inputs of Al to surface waters are probably low
(Lydersen et al., 2002). Acidification has caused a larger input of Al
to the surface waters from the surrounding soils, especially at areas
with low weathering rates in the catchment. Therefore the
concentrations of Al are quite high in natural waters in Sweden.
Similarly as for iron(III), aluminium(III) can precipitate as mineral
phases and be complexed to DOM, but the details are uncertain.
Above pH 6 aluminium solubility is low in both surface waters and
soils, and it is controlled by a mineral phase with Al(OH)3-type
solubility (Tipping, 2005). The actual mineral form that is
controlling Al solubility is not confirmed yet. Suggestions in soils
are amorphous Al(OH)3, gibbsite (crystalline Al(OH)3), and silicacontaining (proto)-imogolite or allophane (Gustafsson et al., 2001).
The theoretical Al(OH)3 phase has a higher log solubility product
than Fe(OH)3, at 25° C it is estimated to be: 8.5 in natural waters
(Tipping, 2005), 8.77 in lakes in the north-east U.S. (Warby et al.,
2008), and 8.29 in moderately acid Bs horizons of podzolised soils
(Gustafsson et al., 2001). Temperature has a large effect on the
Al(OH)3 solubility, the colder the higher solubility product. When
estimating the solubility of Al in soils, Gustafsson et al. (2001) used
9
Carin Sjöstedt
TRITA LWR PHD 1065
tKH KHDW RI UHDFWLRQ ƅHr0, determined for gibbsite as
-105.0 kJ mol-1 (Palmer and Wesolowski, 1992) whereas Tipping
(2005) used a mid-range value from several studies of -107 kJ mol-1
(see Tipping et al., 2002, and references therein). In nature with
high NOM concentrations and pH below 6, Al is often
undersaturated with respect to Al(OH)3 and instead controlled by
complexation to organic matter (Tipping, 2005;; Warby et al., 2008).
The complexation of Al to organic acids is important for
mitigating Al toxicity. However, aluminium has also been known
to compete with other metals for binding sites;; e.g. with Cu
(Tipping et al., 2002;; van Schaik et al., 2010), Eu (Susetyo et al.,
1990;; Bidoglio et al., 1991), Pb (Mota et al., 1996;; Pinheiro et al.,
2000), and Cd (Pinheiro et al., 2000). Aluminium complexation to
organic acids is therefore not only decreasing its own
bioavailability, it could also increase the bioavailability of other
potentially toxic metals.
Aluminium in boreal surface water has been found to be mainly
associated with iron colloids at pH above 6, and not with organic
carbon (Pokrovsky and Schott, 2002;; Vasyukova et al., 2010).
However, for two organic-rich rivers Al was associated with
< 10 kDa organic complexes (Pokrovsky and Schott, 2002).
Aluminium correlated better with organic carbon (OC) than with
Fe for water draining granitic rocks compared to basic rocks
(Vasyukova et al., 2010). In another study with Swedish creek water
(pH 6.5) Al was associated with both iron and organic carbon
colloids (Lyvén et al., 2003).
The inorganic complexes of aluminium include the hydroxidecomplexes, but aluminium also forms inorganic complexes with in
particular F- but also SO42-, and in small amounts SiO4-, CO32- and
Cl-. According to the FIAM, complexation with fluoride would
reduce toxicity of Al due to less Al3+ activity. Lowered toxicity has
been shown in a study of fish, but not to as much as expected due
to the lower Al3+ (Wilkinson et al., 1990). This could be due to
toxicity of Al-fluoride complexes or that the Al-fluoride complexes
are not strong enough to compete with binding sites to the gills
(Gensemer and Playle, 1999).
Inorganic monomeric aluminium (Ali) has often been monitored
on a large scale due to its toxicity at high concentrations. Inorganic
monomeric aluminium can be analysed with several different
methods (Barnes, 1975;; Clarke et al., 1992), but the most common
one is the cation exchange column method developed by Driscoll
(1984). The method retains the inorganic cations, while the anionic
or neutral forms of organic and inorganic Al passes through the
column. The inorganic monomeric Al is determined indirectly as
the difference between total (monomeric) Al and Al in the eluate.
Aluminium can be detected in several different ways, e.g. by
complexation with pyrochatechol violet (PCV) with
spectrophotometric detection. Another more inclusive detection
method uses detection with inductively coupled plasma-optical
emission spectroscopy (ICP-OES). One possible problem for the
10
Iron and aluminium speciation, Implications for geochemical modelling
Driscoll method could be decomplexation of Al from organic
matter when passing through the column, which would
overestimate the inorganic monomeric Al (Backes and Tipping,
1987). Another possible artefact is physical retention of
particulate/colloidal inorganic Al due to its large size in the
column. A problem with all methods is that they generally
overestimate soluble, inorganic monomeric Al. This is due to the
equilibrium shift that occurs when a chelating agent, complexing
agent or a resin is added that releases more Al3+ to the system. The
increased Al3+ is then detected by the complexing agent (Lindsay
and Walthall, 1996;; Gensemer and Playle, 1999).
Many studies have attempted to model Ali in surface waters
(Schecher and Driscoll, 1988;; Driscoll et al., 1994;; Cory et al., 2007;;
Tipping and Carter, 2011). Often it is only inorganic aluminium
that has been modelled. Few studies have modelled both pH and
Ali simultaneously (Driscoll et al., 1994), even though pH and Ali is
closely linked.
This study will focus on the derivation of a model for predicting
both pH and Ali for Swedish surface waters, using a consistent set
of parameters. It is explored how well the model works for
simulating Ali, as determined by the cation-exchange method using
either the PCV or the ICP-OES method for detection. Colloidal
and particulate forms of Al will be studied to see if they relate to a
potential precipitated phase of Al.
1.2.3. Other trace metals
Aluminium is the potentially toxic metal present in the highest
concentrations in Swedish surface water compared to potentially
toxic levels (Lydersen et al., 2002). However, if aluminium is
present together with other metals with lower concentrations it can
result in an additive toxic effect (e.g. Hutchinson and Sprague,
1986). Therefore, it is important to study the speciation of the total
amount of potentially toxic metals, even though they individually
may be present in concentrations that are considered non-toxic.
Copper
Copper is an essential element for all organisms, but it is toxic at
high concentrations, especially to algae, fungi and vertebrates
(Lydersen et al., 2002).
Copper is mainly present in the environment as Cu(I) or Cu(II),
but Cu(I) is only stable at highly reducing conditions. Copper (II)
readily adsorbs to organic matter. Copper has been showed to be
strongly complexed to organic matter in surface waters (Sigg et al.,
2006). In the ultrafiltrated waters of Pokrovsky and Schott (2002)
copper was not appearing as organic or iron-rich colloids, instead
it was mainly in the < 1 kDa fraction (possibly associated with
< 1 kDa organic matter or present as dissolved inorganic species).
Copper in a Swedish creek was associated with organic colloids,
and not with the larger iron colloids (Lyvén et al., 2003).
Copper(II) adsorbs to ferrihydrite, and evidence exists that it
forms inner-sphere edge-sharing bidentate complexes (Scheinost et
11
Carin Sjöstedt
TRITA LWR PHD 1065
al., 2001). Copper(II) might adsorb more strongly in the presence
of other anions, but such systems have rarely been studied.
In this study copper adsorption to ferrihydrite with and without
the presence of phosphate was studied both qualitatively using
EXAFS spectroscopy and quantitatively using batch experiments.
The results have been used to optimise surface complexation
constants using the Charge Distribution Multi Site Complexation
(CD-MUSIC) model (Hiemstra and van Riemsdijk, 1996). Further,
with the new constants Cu has been modelled using a geochemical
equilibrium model in two lakes and compared with ultrafiltration
fractionation to see if the model captures Cu speciation correctly.
Lead
Lead is naturally present in low concentrations. During the 1970s
and 1980s leaded petrol caused widespread pollution of lead to the
environment. As the evidence concerning the potential toxicity of
lead grew, it was phased out from the use as a petrol additive, and
since the 1980s the concentrations have decreased in the
environment (Lydersen et al., 2002).
Lead is mainly present as inorganic Pb(II) in the environment, but
there are also organic lead compounds from human
contamination. Inorganic lead(II) readily adsorbs to iron
(hydr)oxides and complexes to organic matter, and is therefore
often retained in the soil very strongly. Lead is also often
associated with iron and organic matter in natural waters (Fuller et
al., 1988;; Erel and Morgan, 1992;; Lyvén et al., 2003;; Wällstedt et al.,
2009). Lead has been found to be associated with iron colloids and
not with organic colloids in surface water (Pokrovsky and Schott.,
2002;; Lyvén et al., 2003;; Sigg et al., 2006;; Vasyukova et al., 2010).
There may be competition between iron(III) and aluminium(III)
with lead(II) for complexation to organic matter (see sections 1.2.1
and 1.2.2).
Lead adsorption to ferrihydrite has been studied previously using
batch experiments by e.g. Swedlund et al. (2003) and by Gustafsson
et al. (2011). X-ray absorption spectroscopy has been used to study
the complex qualitatively, and both inner-sphere edge-sharing
bidentate complexes (Scheinost et al., 2001;; Trivedi et al., 2003) and
inner-sphere monodentate complexes have been found (Trivedi et
al., 2003). There has been evidence of increased Pb adsorption in
the presence of phosphate on iron oxides (Xie and Giammar,
2007).
Therefore, similar as for Cu, lead adsorption to ferrihydrite with
and without phosphate was examined in this study and the CDMUSIC model was constrained. The resulting CD-MUSIC model
was used together with a consistent geochemical model for
comparison with ultrafiltration fractionation to see if the model
captures Pb speciation correctly for two lakes.
12
Iron and aluminium speciation, Implications for geochemical modelling
1.2.4. Natural organic matter
Natural organic matter (NOM) is a large, heterogeneous group of
compounds. It consists of partly degraded plant, animal and
microbial residues. Parts of it include smaller organic acids such as
citrate, oxalate, alate, acetate and salicylate acids, sugar acids and
phenols. They are easily degraded and therefore not present in
high amounts. The major part of NOM is called humic substances
and they have traditionally been separated into three groups based
upon pH solubility: humin, humic acids and fulvic acids. They are
important for metals since they can complex them and thereby
reduce their bioavailability. It has also been shown that DOM is
not toxic in itself to fish (31 mg L-1 DOC, pH 7) (Richards et al.,
1999). One problem with humic substances is to find a good way
to analyse them (for a review, see e.g. Hedges et al., 2000). The
origin (e.g. from the soil or lake) of the humic and fulvic acids is
not considered important for their characteristics.
Humic and fulvic acids can be separated through acid and base
treatments, and can thereafter be used in titration experiments to
determine the complexation constants for H+ and metal ions. A
question that has been raised is whether the isolation method is
destructive so that the results are not representative for field
conditions. The complexation constants can later be used in
geochemical models. One problem when modelling acid
dissociation is the fact that the acids dissociate over a large range
of pH values due to the heterogeneity of the compounds. Some
patterns do exist. There is a large number of acid groups having
pKa-values between 4 and 6;; these are considered to be carboxylic
acid groups. For another type of group the pKa-values range from
8 to 10;; these are considered to be phenolic acid groups. The
carboxylic acid groups are more involved in complexation of H+
and cations at natural pH values than the phenolic groups.
There are different models available for simulating the proton and
metal binding properties of humic substances. One of the simplest
ones is the equation by Oliver et al. (1983), which is an empirical
model designed for estimating the organic acidity in acidic waters.
Then there are also monoprotic, diprotic and triprotic models
(Driscoll et al., 1994;; Schecher and Driscoll, 1995;; Köhler et al.,
2000;; +UXäka et al., 2003). The most advanced models are the
Non-Ideal Competitive Adsorption Donnan (NICA-Donnan)
model (Kinniburgh et al., 1999), Model VI (Tipping, 1998) with its
precursor Model V (Tipping and Hurley, 1992), and the Stockholm
Humic Model (SHM) (Gustafsson, 2001). These last models
separate between metals bound specifically to the organic ligand
forming complexes and metals accumulated electrostatically due to
the negatively charged ligand.
NICA-Donnan is a model based upon a bimodal, continuous
distribution of complexation sites, divided into the two different
forms of sites: carboxylic and phenolic.
Contrary to the NICA-Donnan model, the Humic binding Model
VI is a model based on discrete sites for binding of protons and
13
Carin Sjöstedt
TRITA LWR PHD 1065
metals and their first hydrolysis product. The Stockholm Humic
Model (SHM) is based upon the Model VI in this respect. The
dissociation reaction is defined as:
ROH ÅÆ RO- + H+, Ki
(2)
where R represents the humic molecule and Ki is an intrinsic
dissociation constant, which includes a correction term for
electrostatics. There are eight sites divided into two groups. Group
A representing strong H+ binding (numbers 1-4), mainly carboxylic
acid groups, and group B with weaker sites (numbers 5-8), mainly
phenolic acid groups. Both groups are represented by four sites
constants each, which are calculated as
(2i 5)
(3)
i 1 4 : log K i log K A
pK A
6
(2i 13)
(4)
i 5 8 : log K i log K B
pK B
6
where log KA, log KB, ƅSKA DQG ƅSKB are needed to define the
eight log Ki values.
The metals can be complexed to the same sites in
monodentate and bidentate complexes in the Stockholm Humic
Model. The monodentate reaction for Cd2+ is defined as:
ROH + Cd2+ ÅÆ ROCd+ + H+, KCdm
(5)
where the constant KCdm includes electrostatics. The same KCdm is
used for the eight proton-binding sites, but the heterogeneity of
site affinity is accounted for by DKHWHURJHQHLW\SDUDPHWHUƅLK2, in
the expression:
log KCdm,x = log KCdm + x · ƅLK2
;; x = 0, 1, 2
(6)
7KH ODUJHU ƅLK2, the more heterogeneity, i.e. a small amount of
sites binds the metal very strongly compared to the majority of
sites. Each site is subdivided into three subsites in the generic
version of SHM, where x is set to 0 for 90.1 % of the sites, 1 for
9 % and 2 for 0.9 % of the sites.
All models assume that the DOM resembles the isolated humic
and fulvic acids. However, generally there can be a fraction of the
DOM that is not humic and fulvic acids, e.g. cellulose, which does
not bind protons or metals. Therefore, there is a need to determine
how much of the natural organic matter that consists of humic and
fulvic acids, i.e. is active in proton and metal binding. Total organic
matter is often estimated by determining the organic carbon
content and multiplying by two, since organic matter generally
consists of approximately 50 % of carbon by weight. The
maximum active fraction of dissolved organic matter compared to
DOC (the so-called ADOM/DOC ratio) is therefore 2, but in
natural waters it is often smaller. Often the ADOM/DOC ratio is
14
Iron and aluminium speciation, Implications for geochemical modelling
the parameter that is fitted when modelling a data set and different
values have been used ranging between 0.6 and 1.98 for surface
waters (Bryan et al., 2002: 1.12-1.98;; Tipping et al., 2002: 1.2-1.4;;
Tipping and Carter, 2011: 0.6-1.35). This is quite a large range and
it has significant importance for pH modelling and metal
speciation. Few have set the fraction to a common value for two
separate parameters, e.g. pH and a metal complexation reaction
(Driscoll et al., 1994).
Recently, an increase of DOC in surface waters in Europe and
North America has been detected (e.g. Driscoll et al, 2003;; Evans et
al. 2005), but the mechanism is under debate.
1.2.5. Mineral surfaces and surface complexation models
Mineral surfaces are important for the adsorption of trace metals.
Iron, aluminium and manganese (hydr)oxides are important
surfaces due to their large abundance in natural waters and their
high amounts of surface groups available for binding. The most
data for metal adsorption exist for the iron (hydr)oxides
ferrihydrite and goethite. Dzombak and Morel (1990) were
pioneers in establishing a database for adsorption to hydrous ferric
oxides and also developed the Diffuse Layer model (which they
referred to as the Generalized Two-Layer model).
The Al, Fe and Mn (hydr)oxides have variable charge depending
on pH. Ferrihydrite has a point-of-zero charge at between 7.9 and
8.3 meaning that at natural pH a majority of the sites are positively
charged (Stachowicz, 2007). This means that it can easily adsorb
inorganic anions and organic acids. Ferrihydrite can also adsorb
metal cations forming strong surface complexes. Ferrihydrite
contains singly-, doubly and triply coordinated oxygen surface
groups, but the singly coordinated ones are the most common and
therefore probably most important for proton binding.
Modelling of adsorption to (hydr)oxides are generally performed
with surface complexation models. They involve surface reactions
treated as complexation reactions in which inner-sphere ions are
adsorbed to discrete sites on the mineral and a mass-action
equation defines the sorption constant K. They also often involve
electrostatics due to counter-ion accumulation in the so-called
diffuse double layer (DDL). The location of the ion charge in the
diffusive double layer is important and will affect the proton coadsorption and release. With spectroscopic data, obtained by e.g.
EXAFS- or infrared (IR) spectroscopy, the structure of the surface
complexes can be determined. By using such data the description
of the sites can become more precise and more realistic. This idea
led to the development of the Multi Site Complexation (MUSIC)
model (Hiemstra et al., 1989a and b). Additionally, it was suggested
that the inner-sphere-adsorbed ions should not be treated as point
charges, since they rather have a distributed charge (Hiemstra and
van Riemsdijk, 1996). This idea is the key concept in the charge
distribution (CD) model (Hiemstra and van Riemsdijk, 1996),
which, in combination with the MUSIC model becomes the CDMUSIC model.
15
Carin Sjöstedt
TRITA LWR PHD 1065
2. M ATERIAL
AND
M ETHODS
2.1. Site description and sampling
Two different areas, Dalsland and Tyresta, where sampled in this
study.
2.1.1. Dalsland - (Paper I)
Three lakes were studied in south-western Sweden, two that were
excessively limed, L. Motjärn and L. Stora Vrångstjärnet in
Dalsland, and one unlimed reference lake, L. Rotehogstjärnen in
Bohuslän. L. Motjärn is 11.3 ha with an average pH of 7.4 and the
maximum depth is 10 m. Lake St Vrångstjärnet is 9.4 ha, with an
average pH of 7.3, and the maximum depth is 12 m. Lake
Rotehogstjärnen is 16.8 ha, with an average pH of 5.3, and with a
maximum
depth
of
10 m
(ISELAW,
2007,
http://info1.ma.slu.se/IKEU). All the lakes are oligotrophic and
situated in boreal forests. The lakes were sampled in September
2007. The lake water chemistry was studied with in-situ dialysis and
geochemical modelling.
2.1.2. Tyresta ² (Paper III and Paper IV)
Tyresta National Park is located 20 km SE Stockholm, Sweden. It
is a boreal forest landscape with thin soil and slowly weathering
bedrock, mainly gneissic-granite (Wällstedt et al., 2009). Four lakes
present in this area have been monitored since 1977. The lakes are
all oligotrophic and dimictic. Two of them, the previously limed L.
Trehörningen and the unlimed L. Årsjön, were sampled in this
study and characterised in detail with regards to water chemistry.
Lake Trehörningen was also modelled for termination of liming in
Paper IV using data from monitoring (see Wällstedt et al., 2009 for
details). One of the other two lakes, L. Långsjön (limed until
1995), was modelled for termination of liming in Paper IV, also
using monitoring data.
Lake Trehörningen was sampled in March 2008, October 2008,
April 2009, January 2010, and March 2011 whereas L. Årsjön was
sampled the same occasions except the first one. Water was taken
from the surface (0.2 m depth) directly into 10-L HDPE
containers that had been acid leached and rinsed with deionised
(Milli-Q) water. The water was further filtrated and run on an
anion column as described in sections 2.2.2 to 2.2.3 and in the
generalised scheme (Fig. 2). General water chemistry and Fe(II)
concentration was also determined.
2.2. Laboratory methods
2.2.1. Dialysis ² (Paper I)
Dialysis is a method that separates molecules based upon size. In
this work a 1 kDa pore size was used to separate small molecules,
mainly inorganic ions and small organic acids, from colloids and
particles. Spectra/por ® regenerated cellulose (Spectrumlabs)
dialysis bags were placed in-situ at the surface and the bottom of
the Dalsland lakes and left for five days to equilibrate. For more
details, see Material and Methods section in Paper I.
16
Iron and aluminium speciation, Implications for geochemical modelling
2.2.2. In-line prefiltration ² (Paper III)
In-line prefiltration was performed with a Geotech 142 mm
polycarbonate (or acrylic) filter holder and with 0.45 µm
polyvinylidene filters. In-line filtration was used to be able to
collect the particles and characterise them using Fe K-edge EXAFS
spectroscopy. Samples from Tyresta were filtrated immediately
upon returning to the lab, starting 3-8 hours after sampling and
finishing within additional 3 hours.
2.2.3. Cross-flow ultrafiltration ² (Paper III)
In cross-flow filtration (CFF) (also called tangential flow filtration)
the water flows fast over a filter and back to a container. Molecules
smaller than the pore size passes through the filter into a container
(the permeate) while larger molecules flows over the filter and
concentrates into a retentate. In this study two 1 kDa regenerated
cellulose membranes (Pellicon2, Millipore), a Pellicon cassette
acrylic filter holder, and a Millipore Masterflex I/P Easyload
peristaltic pump were used. The CF-ultrafiltration generally started
within 30 hours after sampling on the pre-filtered water, generally
it took 2.5-3 hours to filtrate 10 litres. Approximately 10-20 litres
of water were ultrafiltrated with a concentration factor of 20,
resulting in a retentate of approximately half a litre. The retentate
L. Trehörningen and L. Årsjön
In-line 0.45 µm prefiltration
Particles > 0.45 µm
Prefiltered water < 0.45 m
CFF ultrafiltration 1 kDa
AGMP-1 anion column
Retentate > 1 kDa
Freeze-dried retentate
Column residues
Permeate
< 1 kDa
Eluate
Fig. 2. General chart of filtration and preparation of the
water from L. Trehörningen and L. Årsjön. Text in bold
refers to the techniques used. Text in italics refers to the
samples measured by EXAFS spectroscopy. Water
chemistry was characterised for samples in normal font.
17
Carin Sjöstedt
TRITA LWR PHD 1065
was analysed for water chemistry and afterwards freeze-dried. The
freeze-dried retentate was characterised by Fe K-edge EXAFS
spectroscopy. The permeate was analysed for water chemistry and
cation exchange Al fractionation (Driscoll, 1984).
2.2.4. Ferrozine method ² (Paper III)
The first step in the method described in Viollier et al. (2000) was
used to determine iron(II) in the lake water in the field in Tyresta
and for some of the treated water. Ferrozine forms a purple
complex with iron(II) with a maximum absorbance at 562 nm.
Iron(III) also forms a complex with ferrozine but it absorbs much
less at this wavelength and it can be compensated for. The samples
were first pre-filtered with 0.2 µm syringes and then ferrozine was
added and the complex formed was immediately detected with a
spectrophotometer. The samples were also measured without
ferrozine addition to see the contribution to the absorption from
organic matter, and these results were deducted from the samples.
Total Fe concentration in the water was determined by ICP and
finally the iron(III) and iron(II) concentration could be calculated.
2.3. New method for concentrating of iron species ² Anionexchange column
To be able to study the Fe speciation with EXAFS spectroscopy, a
concentration method was needed. In addition to freeze-dried
retentates from CF ultrafiltration (see section 2.2.3) a new method
was developed in Paper III using a column with AGMP-1 anion
exchange resin (Bio-rad, 100-200 mesh). The AGMP-1 resin
consists of strongly basic trimethylammonium groups with a
macro-porous structure composed of a styrene divinylbenzene
copolymer. It adsorbs the negatively charged organic material due
to charge-based separation. The volume of the resin was 3 ml
placed in a 5 ml glass column and the water was pumped with a
flow rate of 1 ml min-1. The residues collected on top of the
column were analysed by EXAFS spectroscopy. The eluate was
monitored for pH, total iron, DOC or absorbance. A test was also
performed determining iron(II) prior and after passing the column.
2.4. EXAFS spectroscopy ² (Paper III and Paper V)
2.4.1. General
X-ray absorption spectroscopy (XAS) is a technique that uses Xrays to characterise the structure around a specific element, for
example coordination atoms and bond lengths. The experiments
are performed at large synchrotrons where charged particles are
accelerated into a booster ring which emits X-rays. From that ring
X-rays can be withdrawn in a so-called beamline, where the
experiments are performed. As for the commonplace use of Xrays in medicine, heavier elements, such as Ca, more easily absorb
the X-rays compared to softer elements, such as C and O.
Element-specific energies are used for exciting an electron from an
inner-electron shell of the atom studied out to continuum (Kelly et
al., 2008). Extra energy causes the electron to move in a wave that
is scattered to other neighbouring atoms. The scattering gives rise
18
Iron and aluminium speciation, Implications for geochemical modelling
1.5
Background function
Data
X-ray absorption
1
0.5
EXAFS region
Absorption edge
0
-0.5
-1
-1.5
6900
7100
7300
7500
7700
7900
Indicent X-ray energy (eV)
Fig. 3. An example of an iron K-edge X-ray absorption spectrum
(= hematite) with the absorption edge and the EXAFS region noted. The
background function (pink) is subtracted from the recorded spectrum to
receive the EXAFS spectrum. eV = electronvolts
to a certain absorption spectrum whenever the waves are in phase
or out of phase (Fig. 3). At certain energies the wave is in phase
due to a certain distance to and to the atomic weight of the
neighbouring atom. The spectrum can therefore be used to
determine the type and the coordination of neighbouring atoms,
including the bond length. In Extended X-ray Absorption fine
structure (EXAFS) spectroscopy uses, the spectrum recorded at
energies above the absorption edge is analysed (Fig. 3).
The iron K-edge EXAFS experiments were performed at MAXLab (MAX II), Lund University at beamlime I-811. Samples (all
from Tyresta) included particles collected from the in-line
prefiltration, colloids from the anion-exchange column, and
colloids from freeze-dried retentates. In April 2009 soil samples
from the Oe horizon surrounding L. Trehörningen were also
included. To improve EXAFS spectra soil subsamples were treated
with additions of 15 mmol kg-1 soil Fe(NO3)3 and pH was adjusted
to 2.6 and 4.0 with HCl and NaOH respectively.
The structure of the lead and copper surface complexes on
ferrihydrite was examined at beamline 4-1 at the Stanford
Synchrotron Radiation Lightsource (SSRL), Stanford, California.
Details are presented in Paper V.
2.4.2. EXAFS Data Analysis
The resulting absorption spectrum has to be treated by first
increasing the resolution by drawing a so-called spline/background
19
Carin Sjöstedt
TRITA LWR PHD 1065
function in the middle of the absorption spectrum (Fig. 3). Later
the background function is subtracted from the absorption
spectrum resulting in an EXAFS spectrum. This spectrum is fitted
with a model representing a possible phase or a mixture present in
the samples. Sometimes different models fit equally well. EXAFS
spectroscopy is often regarded as a method that can falsify a
hypothesis, not verify it.
EXAFS data treatment and modelling of the spectra were
performed using the Athena/Artemis program package (Ravel and
Newville, 2005) which incorporate the FEFF code (Rehr and
Albers, 2000) and Atoms (Ravel, 2001). Wavelet transform (WT)
analysis of EXAFS spectra can be used to differentiate between
heavy and light backscatters for higher coordination shells
(Karlsson et al., 2008). The Morlet wavelet transform incorporated
in the Igor Pro script (Wavelet2.ipf) developed by Chukalina
(2010) was used. The k3-weighted EXAFS spectra were used with
DIUHTXHQF\RIWKHƪSDUDPHWHUIRUDKLJKUHVROXWLRQSORWVIRU
the second shell (2-4 Å), while the half width of the Gaussian
HQYHORSHƳZDVVHWWR
For more details, the reader is referred to paper III and paper V.
2.5. Geochemical modelling ² (Paper I, II, III, IV, V)
2.5.1. Data sets for modelling ² (Paper II and IV)
Large data sets with measured surface water chemistry from
various parts all over Sweden were used to constrain a geochemical
model to be able to simulate pH and Ali simultaneously (Paper II).
In total the different datasets consisted of 5 172 samples for pH
and 3 918 samples for aluminium fractionation. First one data set
was used for optimisation of the ADOM/DOC ratio based on pH
simulations, with samples from a countrywide lake sampling of
unlimed lakes ´0nOVM|DUµn = 322). The ADOM/DOC ratio was
optimised using different values and the pH value was modelled
from charge balance. Subsequently the same data was used to
analyse the usefulness of the default binding constant for
aluminium to predict the measured concentration of inorganic
aluminium and to study the outcome of varying that constant by a
factor of 2.
When evaluating the modelling of the inorganic monomeric
aluminium, data sets with two methods for determining Ali were
used, the PCV-method (n = 2 611) and the ICP-OES method (n =
1307) (section 1.2.2 and Table 1). Due to the different results for
the two methods, the data sets Målsjöar anG ´)RUHVW VWUHDPVµ
(Löfgren et al., 2010) were used as calibration data sets of a
correction function (see section 3.5.2).
One data set consisting of a group of northern streams,
´.U\FNODQµ (Buffam et al., 2007;; Björkvald et al., 2008), had
determined concentrations for both unfiltered and 0.4 µm-filtered
Al and Fe. Measured chemistry using unfiltered concentrations was
modelled for precipitated Al(OH)3 (Table 1) and ferrihydrite, and
20
Iron and aluminium speciation, Implications for geochemical modelling
the modelled precipitates were compared to determined Al and Fe
particles (unfiltered minus filtered concentrations).
The dataset Målsjöar was part a larger national survey of 1 807
unlimed lakes and the 3 043 limed lakes in Sweden (Fölster et al.,
2011a), sampled both in autumn 2007 and spring 2008. These
lakes were simulated for pH as a test of the model. The 3 043
limed lakes were also simulated for lower concentrations of Ca and
Mg, as estimated for termination of liming (Paper IV).
2.5.2. Modelling methods
There are several different geochemical models developed for
metal speciation, like PHREEQC (Parkhurst and Appelo, 1999),
WHAM (Tipping, 1994), Visual MINTEQ (Gustafsson, 2012),
ALCHEMI (Schecher and Driscoll, 1995), MINEQL+
(www.mineql.com) etc. The equations governing inorganic
speciation are generally the same or similar in the models.
The geochemical program Visual MINTEQ (versions 2.53, 2.60
and 3.0) (Gustafsson, 2012) was used for the geochemical
modelling. For inorganic complexes, the thermodynamic default
database was used, which is mostly based on the NIST
compilation (Smith et al., 2003). Total concentrations of analysed
cations and anions were used in the modelling, unless otherwise
stated.
Total iron(II) and iron(III) as determined by the ferrozine method
and ICP-(MS or OES) were entered, either estimated (Paper I) or
determined using the ferrozine method (Paper III). For the
simulations in paper II and IV no measurements of iron(II) were
available, so instead all iron was considered to be iron(III).
If the solubility product was exceeded, the minerals Al(OH)3,
ferrihydrite and for paper I birnessite (MnO2) were allowed to
precipitate. Different log *Ks0 were tested for Al(OH)3 (around
8.29) and ferrihydrite (between 2.69 and 3.8) in paper II. For
birnessite the log *Ks0 value of 18.091 was chosen (Smith et al.,
2003).
The temperature can affect the solubility products Al(OH)3 and
ferrihydrite significantly. The field temperatures for the Tyresta
lakes ranged from ²0.7 to 12.9 ºC, but the pH determination and
aluminium fractionation was performed at room temperature.
Filtration was performed on semi-cold water, in room temperature
but with cold water from the field or from the cold storage room
(4 °C). For the simulations in Paper I and Paper III the measured
field temperature was used. For the data in paper II the field
temperature averaged 7.8 °C and ranged between -1 to 26 °C. For
the data in paper IV the average field temperature was 11.7 °C and
ranged from -1.8 to 24.4 °C. However, since pH, Ali fractionations
and filtration were performed at room temperature an intermediate
value of 10 °C was chosen for paper II and IV.
21
Carin Sjöstedt
TRITA LWR PHD 1065
Table 1. Simplified sketch of aluminium fractionation methods and aluminium
models used. Bold text refers to measured data, normal formatting to fractions
calculated by difference. CExch = cation exchange. PCV = Pyrochatechol violet
Ali = Inorganic monomeric aluminium
METHOD
Total aluminium, HNO3-digested, determined by ICP-OES (Al tot)
PCV
H2SO4-digested, determined by PCV (Al acid sol)
Acid soluble
Not acid-treated, determined by PCV (Al tot mon)
CExch treated (Al org)
ICP-OES
Retained by CExch (Ali)
Not acid-treated, determined by ICP-OES (Al tot mon)
CExch treated (Al org)
FILTRATION
Particles
> 0.45 m
MODEL
Al(OH)3 (s)
Colloids
< 0.45 m > 1 kDa
Organic aluminium
Retained by CExch (Ali)
Permeate
<1 kDa
Inorganic mon. aluminium (Ali)
For dissolved inorganic carbon (IC), measured IC or alkalinity can
be used. However, when measured alkalinity values were used it
occasionally gave erroneous results at high pH values. Instead the
relationship for pCO2 suggested by Sobek et al. (2003) was tested:
pCO2 = (1.079 · DOC + 2.332) · 10-4
(7)
where pCO2 represents the partial pressure of CO2 in atm, and
DOC represents dissolved organic carbon in mg L-1. This
relationship for pCO2 was used in paper II, III, and IV.
Dissolved organic matter was modelled with the Stockholm
Humic Model, with default values for acid-base and metal-binding
parameters. In Fennoscandia little of the total organic carbon
(TOC) analysed in surface waters has been found to be present as
particles larger than 0.45 µm (less than 5 %, Temnerud et al., 2007),
therefore it was assumed that all analysed unfiltered organic carbon
is dissolved organic carbon (DOC). The previous default value of
the active DOM to DOC ratio, 1.4, (derived from the data of
Bryan et al., 2002) was used for the Dalsland lakes (Paper I). For
Paper II this figure was fitted for the pH simulations for the
calibration GDWDVHW ´0nOVM|DUµ DQG GHWHUPLQHG WR be 1.65. This
figure was then used throughout in Paper III and IV. Active DOM
could be humic or fulvic acid, but it was assumed here that all of it
was fulvic acid. Originally SHM contained the dimeric constants of
22
Iron and aluminium speciation, Implications for geochemical modelling
Fe(III) complexation to DOM (Gustafsson et al., 2007). Later on
these were changed in this study to monomeric constants and
fitted for two data sets of mor layers Risbergshöjden Oe and
Korsmossen Oe (Gustafsson et al., 2007) (procedure described in
the appendix of paper II). Complexation constants of Pb to fulvic
acids was also changed according to Gustafsson et al. (2011), and
was used for the modelling of Pb in L. Trehörningen and L.
Årsjön.
Adsorption to ferrihydrite was estimated using the CD-MUSIC
model (Hiemstra and van Riemsdijk, 1996). The model calculated
the amount of adsorbing ferrihydrite that was precipitated from
the total dissolved iron concentration, using the approach
described by Wällstedt et al. (2010). New constants for Cu and Pb
adsorption to ferrihydrite were developed in Paper V.
Complexation constants for Al3+ adsorption to ferrihydrite was
developed in Edkymish (2009), and was only used for modelling
Al in the Tyresta lakes.
2.5.3. Modelling scenarios for termination of liming ² (Paper IV)
To model termination of liming for the 3 043 limed Swedish lakes,
background steady-state concentrations of Ca and Mg not affected
by liming were needed. The method used to estimate these values
was based upon the Ca to Mg ratio in unlimed reference lakes
upstream or within 20 km of the limed lakes. This method relies
on the assumption that Ca and Mg have similar chemical
properties and therefore co-vary in time and space (Fölster and
Wilander, 2005;; Göransson et al., 2006). During liming the Ca to
Mg ratio is increased, as the liming agent has a higher Ca to Mg
ratio than lake water. Therefore to calculate the steady-state Ca
concentration for the unlimed state, the Ca to Mg ratio has to be
adjusted in the limed lake to reach the ratio of the reference lake.
Thus for lakes subject to liming, this principle can be used to
estimate concentrations of Ca after termination of liming (Fölster
et al., 2011b):
2
[Ca ]"unlimed"
2
[Mg ]meas
[Ca 2 ]
[Mg 2 ]
(8)
ref
where [Ca2+]µunlimedµ is the projected steady-state concentration of
Ca after termination of liming, [Mg2+]meas is the measured
concentrations (in the limed lake) and ([Ca2+]/[Mg2+])ref is the ratio
of Ca to Mg in nearby unlimed reference lakes. Detailed
calculations with adjustments for added Mg in the liming agent are
described in Paper IV.
The limed lakes were predicted for pH and Ali DW ´XQOLPHGµ
conditions using these values of Ca and Mg concentrations and the
model described in Paper II. Model predictions of what would
happen if liming was decreased in a step-wise manner was also
performed, with progressively lower and lower concentrations of
Ca and Mg. The concentrations of added Ca and Mg from lime
were decreased from 100 % (determined levels at the present
23
Carin Sjöstedt
TRITA LWR PHD 1065
situation) down to 75 %, 50 %, 25 % and 0 % (calculated
´XQOLPHGµFRQFHQWUDWLRQV
To get evidence for the applicability of the model, six lakes with
monitored termination of liming were studied in the same way as
for the large data set. The month when the Ca and Mg
concentrations would reach these ratios were calculated based
upon lake retention time. A comparison was then made between
the predicted and the analysed values of pH and Ali for that
month.
3. R ESULTS
3.1. Iron: size fractionation and characterisation
Iron was mostly present in the colloidal (>1 kDa, <0.45 µm) and
particulate fractions (>0.45 µm), for both Lake Trehörningen and
Lake Årsjön as well as for the surface water of L. Motjärn, L. St.
Vrångstjärnet and L. Rotehogstjärnen (Fig. 4-13). The colloidal
fraction was often the largest compared to the particle fraction in
the Tyresta lakes. Iron(II) was determined to be present in the
surface water of L Trehörningen and Årsjön, sometimes as high as
34 % of total iron (Fig. 5;; Table 1 in Paper III). The percentage of
iron(II) could not be correlated to any of the parameters pH or
TOC (data not shown). It was also evident that the ratio of iron(II)
to total iron could both increase and decrease rapidly in the
samples at the laboratory, where light was shown to increase the
fraction of iron(II) (data not shown).
Total iron was correlated to OC in the L. Trehörningen and L.
Årsjön fractionation, especially for the smaller fractions (Fig. 14a).
For example, the L. Trehörningen October 2008 sample contained
a large amount of iron particles (566 µg L-1), but very little OC
particles (0.1 mg L-1) (Table 1 in Paper III).
100%
Permeate < 1 kDa
80%
Colloids < 0.45 m >1 kDa
Particles > 0.45 m
60%
Inorganic dissolved
Bound to DOM < 1 kDa
40%
Bound to DOM > 1 kDa
Adsorbed to ferrihydrite
Precipitated
20%
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Cu
filtered
Cu
modelled
Fig. 4. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Trehörningen from the October 2008
sampling. Chemistry of bulk water: pH 5.23, TOC 20.5 mg L-1, Al 680 µg L-1,
Fe 1 380 µg L-1, Pb 1.3 µg L-1 and Cu 0.49 µg L-1.
24
Iron and aluminium speciation, Implications for geochemical modelling
The EXAFS results of iron showed mainly iron(III) in all samples,
as judged by the typical Fe-O distance of 2.0 Å. The L.
Trehörningen particle sample in October 2008 had a distinct
resemblance to ferrihydrite, but with a higher Debye-Waller factor
for Fe-O in the first coordination shell and fewer coordinated iron
and oxygen atoms in higher shells (Table 4 and Fig. 2 in Paper III).
The two techniques for preconcentration of the colloids gave
slightly different results. The anion-exchange column samples
showed iron(III) with Fe···C distances between 2.77 and 2.85 Å
(Table 4 in paper III), interpreted as iron complexed in monomeric
complexes to NOM. In contrast, the freeze-dried retentate
samples showed either a mixture of this monomeric complexation
and an iron (hydr)oxide phase, or one or the other of those two
phases. Three samples were analysed using both preconcentration
methods. In one case both methods gave consistent results, but
for two occasions the retentates showed a higher amount of
polymeric iron. Generally the EXAFS results showed a
predominance of iron(III) complexed to organic matter in the
colloids of the two lakes.
The soil samples contained either iron(III) complexed
monomerically to organic matter, or a mixture of this phase and of
an iron (hydr)oxide phase. The iron (hydr)oxide phase was too low
in concentration to be identified.
The obtained results, showing that monomeric complexation of
Fe(III) to DOM was the predominating complexation mechanism,
motivated a change of the Fe(III) complexation in the SHM model
from the generic dimeric constants to new monomeric ones. The
new complexation constants were optimised as described in the
Methods section and in the Appendix of Paper II and are
presented in Table 2. The monomeric constants resulted in a
stronger complexation to organic matter. When iron speciation
was simulated for L. Trehörningen and L. Årsjön iron(III) was
100%
Permeate < 1 kDa
80%
Colloids < 0.45 m >1 kDa
Particles > 0.45 m
60%
Inorganic dissolved
Bound to DOM < 1 kDa
40%
Bound to DOM > 1 kDa
Adsorbed to ferrihydrite
20%
Precipitated
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Fig. 5. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Trehörningen in the April 2009 sampling.
Chemistry of bulk water: pH 5.25, TOC 20.9 mg L-1, Al 666 µg L-1,
Fe 1 096 µg L-1 where 34 % was Fe(II), and Pb 1.1 µg L-1.
25
Carin Sjöstedt
TRITA LWR PHD 1065
generally found to complex to organic matter instead of
precipitating as ferrihydrite (Fig. 4-10). This was also in accordance
with the EXAFS results. The geochemical modelling also showed
ferrihydrite in approximately the same amount as Fe particles
(Fig. 4-10). This was further studied in the Krycklan data set, see
section 3.5.3.
3.2. Aluminium: size fractionation and modelling in the five lakes
Aluminium was not present in the particulate fraction of L.
Trehörningen and L. Årsjön, other than in very small amounts
(< 13 % of total aluminium). Instead it was present in the colloidal
fraction (between 28 and 56 %) or in the permeate (between 34
and 69 %). In the excessively limed lakes, Al was present mainly in
the colloidal/ particle fraction (between 51 and 82 %) compared to
the dialysed fraction (Fig. 11 and 12). In the acid L.
Rotehogstjärnen aluminium was also present mainly in the
colloidal fraction (75 %) (Fig. 13).
Aluminium, organic carbon and total iron were strongly correlated
to each other in the L. Trehörningen and L. Årsjön fractions
(Fig. 14a and b;; Fig. 15a).
For the two excessively limed lakes, most of the Al was simulated
to be precipitated Al(OH)3, in accordance with the dialysis (Fig. 11
and 12). In contrast, the geochemical model suggested that no
aluminium was precipitated as Al(OH)3 in the Tyresta lakes (Fig. 410) and in L. Rotehogstjärnen (Fig. 13), probably due to the low
pH (below 5.6) and low temperatures. This was consistent with the
low amount of particles. The small amount of particles that existed
was present with a positive relationship to iron particles, but in a
smaller fraction of total Al (Fig. 4-9). They could be modelled with
adsorption to ferrihydrite, albeit in a smaller fraction than
measured (Fig. 4-9).
Much of the Al was modelled to be complexed to organic matter,
for all the samples from L. Trehörningen (Fig. 4-6), L. Årsjön
(Fig. 7-10) and L. Rotehogstjärnen (Fig. 13).
3.3. The speciation of As and Mo in Dalsland ² (Paper I).
Arsenic was present in the < 1 kDa fraction in around 80 % of
total arsenic in the excessively limed lakes and at 69 % in the acid
L. Rotehogstjärnen (Fig. 11-13). Molybdenum was mainly present
in the < 1 kDa fraction (Fig. 11-13). Arsenic, and also to some
extent molybdenum, were influenced of the modelled speciation of
Table 2. Initial generic complexation constants of iron(III) to fulvic acid
(dimeric, tridentate;; Gustafsson et al., 2007) and the new constants for
monomeric, bidentate complexation in the SHM.
New constants for monomeric complex
Dimeric complex
log K
Complex
+
(FA)3Fe2O
-5.15
ǻLK2
a
1.8 /1.5
b
Complex
log K
ǻLK2
(FA)2FeOH
-4.6
1.7
-1.68
1.7
(FA)2Fe
a
+
b
1.8 = fitted for soil organic matter, 1.5 = fitted for humic acids (see Gustafsson et al., 2007 for details)
26
Iron and aluminium speciation, Implications for geochemical modelling
iron, since they both can adsorb to ferrihydrite. Arsenic adsorption
to ferrihydrite was generally exaggerated by the model. The
modelling of molybdenum generally predicted low adsorption to
ferrihydrite and low organic complexation, which coincided with
the measured concentrations (Fig. 11-13).
3.4. Copper(II) and lead(II) adsorption to ferrihydrite, as influenced
by phosphate ² (Paper V)
The adsorption of copper(II) and lead(II) to ferrihydrite was
significantly increased in the presence of phosphate (Fig. 16 and
17). This was especially true for lower pH and more for Pb than
for Cu. For copper(II) a possible picture of the system was
constructed using the EXAFS results and the batch experiments in
combination with CD-MUSIC modelling. In the absence of
phosphate copper was bound in a bidentate edge-sharing complex
to ferrihydrite. However, in the presence of phosphate copper was
likely bound partly in a ternary complex with phosphate where Cu
was monodentately bound to one iron octahedron and at the same
time coordinated to one phosphate. The phosphate was in turn
complexed to another iron octahedron. The constants in the CDMUSIC model were changed accordingly and could thereafter
describe the system well (Fig. 16).
For Pb the results were the same as for Cu without phosphate
present, i.e. it formed a bidentate edge-sharing complex. However,
in the presence of phosphate the results were a bit more
ambiguous. The data could be interpreted in two ways, and both
are equally likely. Either it was similar as for Cu, but with a smaller
angle than 180° for the monodentate binding to ferrihydrite. The
other option is also a ternary complex, but where Pb(II) is bound
bidentately to two corner-sharing iron octahedrons and the
phosphate ion interacts only with Pb(II) and not with the surface.
The constants in the CD-MUSIC model were adjusted according
100%
Permeate < 1 kDa
80%
Colloids < 0.45 m >1 kDa
Particles > 0.45 m
60%
Inorganic dissolved
Bound to DOM < 1 kDa
40%
Bound to DOM > 1 kDa
Adsorbed to ferrihydrite
20%
Precipitated
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Fig. 6. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Trehörningen in the January 2010 sampling.
Chemistry of bulk water: pH 4.88, TOC 22.7 mg L-1, Al 668 µg L-1,
Fe 1 270 µg L-1 where 2 % Fe(II), and Pb 1.3 µg L-1.
27
Carin Sjöstedt
TRITA LWR PHD 1065
100%
80%
Permeate < 1 kDa
Colloids < 0.45 m >1 kDa
60%
Particles > 0.45 m
Inorganic dissolved
40%
Bound to DOM <1 kDa
Bound to DOM >1 kDa
Adsorbed to ferrihydrite
20%
Precipitated
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Cu
filtered
Cu
modelled
Fig. 7. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Årsjön in the October 2008 sampling.
Chemistry of bulk water: pH 5.61, TOC 10.2 mg L-1, Al 184 µg L-1,
Fe 304 µg L-1, Pb 0.28 µg L-1 and Cu 0.53 µg L-1.
to the first alternative, and it fitted the data well (Fig. 17).
However, the adjusted CD-MUSIC model parameters were in
agreement also with what would be expected for the second
alternative complex.
3.5. A consistent geochemical model ² (Paper II and Paper IV)
A consistent geochemical model for Swedish surface water was
derived based upon the results above. The new monomeric
constants for Fe(III)-DOM complexation (see section 3.1) were
used throughout. The new constants for copper and lead sorption
to ferrihydrite were used in section 3.5.4.
3.5.1. Modelling pH
The only factor that was varied in the calibration of the model was
the fraction of active dissolved organic matter to DOC. This was
calibrated to be ADOM/DOC = 1.65 when simulating pH from
the charge balance for the calibration data set of the unlimed
reference lakes (Fig. 18a). Assuming that DOM consists of 50 %
carbon by weight, the value of 1.65 implies that 82.5 % of the
DOM was active with regards to proton and metal binding and the
calculated site density is 11.5 µM mg-1 C. The model was validated
for a large number (n = 4 850) of lakes and rivers in Sweden in
Paper II (Fig. 18b) and 4 528 lakes in paper IV (Fig. 19a and b)
with a mean error of 0.2 pH units. The choice of temperature and
solubility constant for Al(OH)3 and ferrihydrite had minor effects
on the pH and Ali simulations.
3.5.2. Modelling inorganic monomeric aluminium
Ali was simulated with a strong positive relationship for the two
calibration data sets for the detection methods PCV and ICP-OES
(Fig. 20a and b). They differed however slightly from the modelling so a correction function was included in the modelling:
28
Iron and aluminium speciation, Implications for geochemical modelling
Ali * = a + b · Ali (MOD)
(9)
-1
where Ali * is the new corrected concentrations in µg L Al and Ali
(MOD) is the geochemically modelled concentrations of inorganic
monomeric Al and a = 12.4 and b = 1.10 for the PCV method and
a = -4.9 and b = 0.74 for the ICP-OES method (Fig. 20c and d).
After that Ali was validated on a number of datasets, and the mean
error was 30 µg L-1 for the PCV-method (n = 2 289) and 33 µg L-1
for the ICP-OES method (n = 1 130) (Paper II). Samples with pH
values above 6 were excluded from the graphs and in calculations,
due to possible artefacts such as particulate/colloidal inorganic Al
retained in the column.
3.5.3. Modelling aluminium and iron particles
For the Krycklan data set the largest concentrations of particulate
Al and Fe were present above pH 5.5 (data not shown). Modelled
Al(OH)3 was compared to measured Al in particles in Krycklan for
pH values above 5.5. There was a good correlation (Fig. 21a).
Precipitated ferrihydrite was also compared with measured
particles in Krycklan. Here there was also a significant relationship,
but with more ferrihydrite modelled than in measured particles
(Fig. 21b). With a higher log *Ks0 of 3.8 too little ferrihydrite was
precipitated, which is not likely. Different temperatures also
influenced the results. A higher temperature than 10 ºC resulted in
more Al(OH)3 and ferrihydrite precipitating.
3.5.4. Modelling copper and lead fractionation
Copper had a size fractionation that differed a lot between
samplings with no clear pattern to either iron or OC (data not
shown). There were some problems with contamination of Cu
during pre-filtration and ultrafiltration, therefore only
uncontaminated fractionation samples are presented (Fig. 4, 7, and
9). Copper was present in the particulate fraction with a maximum
100%
Permeate < 1 kDa
80%
Colloids < 0.45 m >1 kDa
Particles > 0.45 m
60%
Inorganic dissolved
Bound to DOM <1 kDa
40%
Bound to DOM >1 kDa
Adsorbed to ferrihydrite
Precipitated
20%
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Fig. 8. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Årsjön in the April 2009 sampling. Chemistry
of bulk water: pH 5.53, TOC 10.5 mg L-1, Al 217 µg L-1, Fe 379 µg L-1 where
18 % was Fe(II), and Pb 0.40 µg L-1.
29
Carin Sjöstedt
TRITA LWR PHD 1065
of 20 %. It was also not adsorbed to ferrihydrite, according to the
model. Instead most of it was complexed to DOM (between 74
and 90 %). Cu was also modelled well when simulating the amount
present in the of < 1 kDa fraction for L. Årsjön in October 2008
and January 2010 (Fig. 7 and 9), but it was too low simulated for
the L. Trehörningen October 2008 sample (Fig. 4).
For lead the simulated speciation was closely related to iron
speciation;; if ferrihydrite precipitated, Pb was modelled to adsorb
to it in similar percentages (Fig. 4-9). The low concentrations of
phosphate (generally < 0.1 µM) in these lakes had a small impact
on the adsorption to ferrihydrite according to the new model;;
modelled adsorption was increased by a maximum of five
percentage units when including phosphate as a species. The major
part of Pb was generally modelled to be complexed to DOM
(Fig. 4-9). The modelling results were generally in agreement with
the ultrafiltration results, although with a larger fraction of
< 1 kDa modelled than measured (Fig. 4-9). Lead followed closely
the size fractionation of iron in the L. Trehörningen and L. Årsjön
water (Fig. 15b). Lead was also correlated to the OC size
fractionation (data not shown).
3.6. Predicted changes in lake chemistry at terminated liming ²
(Paper IV)
The model setup was first tested for the six monitored lakes with
terminated liming. For the initial limed situation, there was a very
good correspondence between simulated and analysed pH (mean
error 0.07 pH units, n = 6) (Table A1 in Paper IV). Unfortunately
there was just one lake with analysis results for Ali available before
liming was discontinued. The calculated Ca and Mg concentrations
generally agreed with measured concentrations for the 75, 50, 25
and 0 % lime effect for lakes limed on the lake surface for the
previous 10 years (Fig. 1a and b in Paper IV). More importantly,
100%
80%
Permeate < 1 kDa
Colloids < 0.45 m >1 kDa
60%
Particles > 0.45 m
Inorganic dissolved
Bound to DOM <1 kDa
40%
Bound to DOM >1 kDa
20%
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Pb
filtered
Pb
modelled
Cu
filtered
Cu
modelled
Fig. 9. Size fractionation using pre-filtration and ultrafiltration compared to
modelled metal speciation in L. Årsjön in the January 2010 sampling.
Chemistry of bulk water: pH 5.26, TOC 10.3 mg L-1, Al 182 µg L-1,
Fe 335 µg L-1 where 11 % was Fe(II), Pb 0.35 µg L-1 and Cu 0.87 µg L-1.
30
Iron and aluminium speciation, Implications for geochemical modelling
100%
80%
Permeate < 1 kDa
Colloids < 0.45 m >1 kDa
60%
Particles > 0.45 m
Inorganic dissolved
40%
Bound to DOM <1 kDa
Bound to DOM >1 kDa
20%
0%
TOC
filtered
Al
filtered
Al
modelled
Fe
filtered
Fe
modelled
Fig. 10. Size fractionation using pre-filtration and ultrafiltration compared to modelled metal speciation in L.
Årsjön in March 2011. Chemistry of bulk water: pH 5.18,
TOC 11.8 mg L-1, Al 223 µg L-1, and Fe 414 µg L-1 where 5 %
was Fe(II).
pH and Ali were simulated very well (mean error 0.19 pH units,
n = 8;; mean error Ali 10.6 µg L-1) (Fig. 22a and b;; Table A1 in
Paper IV). However, for lakes with recent upstream liming;; too
low Ca and Mg concentrations were calculated (Fig. 1a and b and
Table A1 in Paper IV). This is not surprising since the influent
water probably had a higher Ca to Mg ratio than the unlimed
reference value. Consequently, the model occasionally predicted
too low pH and too high Ali (mean error 0.66 pH units, n = 11;;
mean error Ali 20.5 µg L-1, n = 8) (Fig. 22a and b, Table A1 in
Paper IV).
When the 3 043 limed Swedish lakes were simulated with the
FDOFXODWHG´XQOLPHGµFRQFHQWUDWLRQVRI&DDQG0Ji.e., the steadystate concentrations after complete termination of liming), pH was
lowered with, on average, 0.92 units compared to the simulated
original situation. For 39 % of the lakes, the pH was calculated to
be below 5.6. Ali was increased with on average 17.9 µg L-1 up to
31.1 µg L-1. Thirty-four per cent of the lakes had Ali concentrations
above 30 µg L-1. However, forty per cent of the lakes maintained a
pH above 6.0, which is above the policy aims for liming in
Sweden. If only the samples from the same lakes taken in the fall
2007 were modelled, still 30 % of the lakes were predicted to have
a pH above 6.0 after terminated liming.
When liming was modelled to decrease to 75 % and 50 % of the
current level, there were small effects on pH and Ali compared to
the limed situation (Table 3 and Fig. 3b and c in Paper IV). Only a
few per cent of the lakes were predicted to have a pH below 5.6
and an Ali above 30 µg L-1.
The decrease in pH and the associated increase in Ali were very
different for different lakes, as the range of individual pH and Ali
changes was 3 pH units and 450 µg L-1 Ali, respectively. There was
31
Carin Sjöstedt
TRITA LWR PHD 1065
a clear geographical pattern, with lower pH and higher Ali in lakes
in south-western Sweden (Fig. 23 a and b), corresponding largely
to historical and present sulphur deposition patterns (Fig. 24 a and
b).
4. D ISCUSSION
4.1. Iron speciation
Iron(II) could be present in relatively large concentrations in the
surface water of the Tyresta lakes, even though the water was fully
oxygenated. This was probably due to high amount of humic
material and the acid pH which can help in keeping Fe(II) in high
steady-state concentrations (Miles and Brezonik, 1981). The
geochemical model was not able to predict the concentration of
iron(II) from available environmental monitoring data. This could
mean that it would be advisable to determine Fe(II) if geochemical
modelling is to be performed, which was also concluded by Lofts
et al. (2008).
According to the EXAFS results, iron was present as monomeric
complexes to DOM and /or an iron (hydr)oxides in the two
Tyresta lakes. Other authors have also found monomeric
complexation of iron(III) to NOM (Karlsson et al., 2008;; van
Schaik et al., 2008 and Karlsson and Persson, 2010). The results
suggested a stronger complexation than previously suggested from
chemical equilibrium modelling. This means that iron(III) could
compete strongly with other metals. The new monomeric
constants made Fe(III) a strong competitor with Al(III), which
resulted in a larger concentration of Ali than was previously
modelled. This was in accordance with the determined Ali
concentrations. None of the samples with organically complexed
iron(III) present showed any evidence for di- or trimerically
complexed iron(III), as reported by Rose et al. (1998) and VilgéRitter et al. (1999), later supported by Gustafsson et al. (2007) and
100%
<1 kDa dialysed
80%
>1 kDa dialysed
Inorganic
60%
Bound to DOM <1 kDa
Bound to DOM >1 kDa
40%
Adsorbed to ferrihydrite
20%
Precipitated
0%
Al
dialysed
Al
modelled
As
dialysed
As
modelled
Fe
dialysed
Fe
modelled
Mo
dialysed
Mo
modelled
Fig. 11. The dialysable (white) and the non-dialysable (black) fractions of the
metals in % of the total concentrations and modelled speciation in the
surface water of L. Motjärn. Chemistry of bulk water: pH 7.69,
TOC 7.1 mg L-1, Al 36.4 µg L-1, As 0.34 µg L-1, Fe 60 µg L-1, and Mo 0.05 µg L-1.
32
Iron and aluminium speciation, Implications for geochemical modelling
100%
<1 kDa dialysable
80%
>1 kDa non-dialysable
Inorganic dissolved
60%
Bound to DOM <1 kDa
40%
Bound to DOM >1 kDa
Adsorbed to ferrihydrite
20%
Precipitated
0%
Al
dialysed
Al
modelled
As
dialysed
As
modelled
Fe
dialysed
Fe
modelled
Mo
dialysed
Mo
modelled
Fig. 12. The dialysable (white) and the non-dialysable (black) fractions of the
metals in % of the total concentrations and modelled speciation in the
surface water of L. St. Vrångstjärnet. Chemistry of bulk water: pH 7.33,
TOC 10.1 mg L-1, Al 112 µg L-1, As 0.35 µg L-1, Fe 79 µg L-1, and Mo 0.04 µg L-1.
by Mikutta and Kretzschmar (2011). It is still not clear which
chemical conditions that favour different binding modes.
Differences in pH, DOM and iron concentrations may be possible
determinants.
Ferrihydrite seemed to be the predominant iron phase in L.
Trehörningen particles, but it did not completely resemble
laboratory phases of ferrihydrite. This could be due to substitution
of other elements into the structure or adsorption of DOM which
could change the size and structure. Another possibility is that
pure ferrihydrite was mixed with e.g. iron(III) complexed to DOM,
since the EXAFS results show an average of all iron present in the
sample. As carbon is lighter than iron it is more difficult to detect
100%
<1 kDa dialysed
80%
>1 kDa dialysed
60%
Inorganic
Bound to DOM
unknow n MW
Adsorbed to ferrihydrite
40%
20%
Precipitated
0%
Al
dialysed
Al
modelled
As
dialysed
As
modelled
Fe
dialysed
Fe
modelled
Mo
dialysed
Mo
modelled
Fig. 13. The dialysable (white) and the non-dialysable (black) fractions of the
metals in % of the total concentrations and modelled speciation in the
surface water of L. Rotehogstjärnen. Chemistry of bulk water: pH 5.54,
TOC 15.5 mg L-1, Al 328.3 µg L-1, As 0.60 µg L-1, Fe 578 µg L-1, and
Mo 0.03 µg L-1.
33
Carin Sjöstedt
TRITA LWR PHD 1065
iron(III) complexation to DOM;; it is masked in the presence of
sufficient amounts of polymerised iron.
The solubility constant of ferrihydrite of log *Ks0 = 2.69 seemed to
fit the data in Tyresta, Dalsland and Krycklan. For Krycklan there
was a higher amount of modelled ferrihydrite than measured
particles (Fig. 21b), but this could be explained by presence of
colloidal ferrihydrite smaller than 0.4 µm. The discrepancies for
Krycklan could also be attributed to dissolved iron(II) not
considered in the model. The choice of temperature influenced the
simulation results regarding how much of the minerals that
precipitated.
4.2. The anion exchange column method
The new method developed in this study for concentrating of iron
species prior to EXAFS spectroscopy, i.e. by use of a column with
1600
900
1400
800
700
1200
600
Al µg L -1
Fe µg L -1
1000
800
600
500
400
300
400
200
200
100
0
0
0
5
10
15
20
25
0
5
10
OC m g L-1
15
20
25
OC m g L-1
Fig. 14a and b. Variations of iron (a) and aluminum (b) versus organic
carbon concentrations in the different filtrates (symbols): Total
concentrations, pre-filtered (> 0.45 µm) and permeate (<1 kDa) at the same
sampling connected by lines for L. Trehörningen (blue diamonds) and L.
Årsjön (red circles).
900
2
800
1.8
700
1.6
1.4
Pb µg L-1
Al µg L-1
600
500
400
300
1.2
1
0.8
0.6
200
0.4
100
0.2
0
0
0
200
400
600
800
Fe µg L-1
1000
1200
1400
0
200
400
600
800
Fe µg L-1
1000
1200
1400
Fig. 15a and b. Variations of aluminium (a) and lead (b) versus iron
concentrations in the different filtrates (symbols): Total concentrations, prefiltered (> 0.45 µm) and permeate (<1 kDa) at the same sampling connected
by lines for Lake Trehörningen (blue diamonds) and L. Årsjön (red circles).
34
Iron and aluminium speciation, Implications for geochemical modelling
the anion-exchange resin AGMP-1, proved to be a good method
for isolating negatively charged iron species. It had a higher
recovery than the freeze-dried retentate and it did not require
changes in temperature and pressure compared to natural
conditions. This would imply that it more closely resembles the
speciation in the lake. From the results of this study it seems that
the preparation of the freeze-dried retentates may have caused
precipitation of an iron (hydr)oxide in some cases, implying that
the anion-exchange column method is a better technique for
preconcentrating the species. The results also showed that it was
important that the isolation was performed in the dark, since in the
presence of light some of the iron(III) was reduced to iron(II) in
the tubing feeding the column;; as a result iron passed through the
column as a free ion (Fe2+).
Generally, rapid oxidation/reduction reactions of iron constitute
an uncertainty when characterising iron speciation. Therefore it is
important to determine Fe(II) in the field as was done in this
study. For samples subject to isolation and characterization using
EXAFS spectroscopy, it is important to avoid speciation changes
as much as possible by storing the samples in the dark and at 5 ºC.
4.3. Aluminium speciation
In Visual MINTEQ simulations, consideration of Al(OH)3
precipitation with a log *Ks0 = 8.29 (9.27 at 10 ºC) seemed to work
for Krycklan and in the Dalsland lakes for pH values above 5.5.
For the Krycklan data set there was a strong relationship between
measured Al in particles and simulated levels of precipitated
Fig. 16. Results from adsorption batch experiments with copper and ferrihydrite (symbols) and geochemical modelling with Visual MINTEQ (lines),
with addition of phosphate (black triangles) and without addition of
phosphate (green squares).
35
Carin Sjöstedt
TRITA LWR PHD 1065
Al(OH)3. Below approximately pH 5.5 Al was undersaturated with
respect to Al(OH)3. The results are similar to those of Tipping
(2005), in which a solubility control by Al(OH)3 with log *Ks0 = 9.5
at 10 ºC could be used for surface waters at pH above 5.8.
However, from the results it cannot be concluded if the mineral
was truly Al(OH)3 or if Si was included, due to an excessive
amount of Si present. Another possibility not tested in Paper I and
II was adsorption of aluminium onto an iron (hydr)oxide. This
may be a possible explanation to the Al present in the particles of
the water from L. Trehörningen and L. Årsjön (Fig. 4-9). Here pH
was too low for precipitated Al(OH)3 to be stable according to the
model. This was also the case for many of the samples of the
larger data sets. Undersaturation of Al(OH)3 has been found
previously by Tipping (2005) and Warby et al. (2008).
The modelling and size fractionation results mainly indicated Al to
be associated with organic colloids, which is in agreement with a
study of Norwegian acid streams, where 45-70 % of the total Al
was above 10 kDa and probably was organic colloids (Teien et al.,
2007).
Fig. 17. Results from adsorption batch experiments with lead and ferrihydrite
(symbols) and geochemical modelling with Visual MINTEQ (lines), with
addition of phosphate (black triangles) and without addition of phosphate
(green squares).
36
8.5
8.5
8
8
7.5
7.5
7
7
pH simulated
pH simulated
Iron and aluminium speciation, Implications for geochemical modelling
6.5
6
6.5
6
5.5
5.5
5
5
4.5
4.5
4
4
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
4
4.5
5
pH analysed
5.5
6
6.5
7
7.5
8
8.5
pH analysed
Fig. 18 a) Simulated S+ LQ WKH FDOLEUDWLRQ GDWD VHW ´0nOVM|DUµ Q = 322,
rmse = 0.19 pH units.
b) Simulated S+ LQ WKH YDOLGDWLRQ GDWD VHW ´'DODUQDµ Q = 126,
rmse = 0.23 pH units.
4.3.1. Aluminium fractionation
The two detection methods of Ali, PCV and ICP-OES, gave
slightly different results as judged by the two calibration data sets
(Fig. 20a and b). The methods are operationally defined and differ
in flow rate (3.8 (ICP-OES) and 2.8 (PCV) mL per mL exchanger
volume) and it is therefore not strange that they do not completely
coincide with each other and with the model. Therefore a
correction equation was introduced (Eq. 8, Fig. 20c and d).
For the data sets in paper II the PCV method generally showed
low concentrations of Ali at pH values above 6, which is expected.
This has not generally been the case for the ICP-OES method,
which might include retained polymeric Al at high pH values
(Andrén and Rydin, 2009). At lower pH and high DOC there may
8.5
8.5
8
8
7.5
7.5
pH simulated
pH simulated
7
6.5
6
7
6.5
6
5.5
5.5
5
5
4.5
4.5
4
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
4.5
5
5.5
6
6.5
7
7.5
8
8.5
pH analysed
pH analysed
Fig. 19 a) Simulated pH in the validation data set of the unlimed reference
lakes spring 2008. n = 1 485, rmse = 0.20 pH units.
b) Simulated pH in the validation data set of the limed lakes spring 2008.
n = 3 043, rmse = 0.20 pH units.
37
Carin Sjöstedt
TRITA LWR PHD 1065
be partial dissociation of Al complexed to DOC for both methods,
erroneously detecting a larger concentration of Ali (Backes and
Tipping, 1987;; Teien et al., 2007).
4.4. Methods for studying trace metal speciation
Dialysis has its advantages due to its in-situ speciation, its low
contamination and its less disruptive method of separation
compared to cross-flow filtration. However, it could not be used
for preconcentrating samples for EXAFS spectroscopy. With
cross-flow filtration possible artefacts include a higher
contamination risk, adsorption to the filters, aggregation or
destroying of colloids. The anion-exchange column method solved
some of these problems, see section 4.2. However, trace metal
speciation is more or less sensitive to temperature changes,
changes in redox conditions, light etc and there is always an
uncertainty when transferring results determined at the laboratory
400
400
350
350
300
Ali simulated (µg L-1 Al)
Ali simulated (µg L-1 Al)
300
250
200
150
100
250
200
150
100
50
50
0
0
50
100
150
200
250
300
350
0
400
0
100
-1
200
300
400
Ali analysed (µg L-1 Al)
400
400
350
350
300
300
Ali * simulated (µg L-1 Al)
Ali * simulated * (µg L-1 Al)
Ali analysed (µg L Al)
250
200
150
250
200
150
100
100
50
50
0
0
0
50
100
150
200
250
300
350
400
Ali analysed (µg L-1 Al)
0
50
100
150
200
250
300
350
400
Ali analysed (µg L-1 Al)
Fig. 20. a) Simulated Ali versus PCV-determined Ali for the calibration data
VHW´0nOVM|DUµUPVH —J/-1 Al. b) Simulated Ali versus ICP-OES-deterPLQHG $OL IRU WKH GDWD VHW ´)RUHVW VWUHDPVµ UPVH = 38 µg L-1 Al. c) The
corrected simulated Ali* using Eq. 9 versus PCV determined Ali for the calLEUDWLRQGDWDVHW´0nOVM|DUµUPVH —J/-1 Al. d) The corrected simulated
Ali* using Eq. 9. versus ICP-OES determined Ali IRU WKH GDWD VHW ´)RUHVW
streamsµ UPVH = 26 µg L-1 Al. Data with pH values above 6 are excluded
from the analysis and this graph.
38
2500
5000
2000
4000
Simulated ferrihydrite (µg L -1 Fe)
Simulated Al(OH)3 (µg L-1 Al)
Iron and aluminium speciation, Implications for geochemical modelling
1500
1000
500
0
-500
-500
0
500
1000
1500
2000
2500
3000
2000
1000
0
-1000
-1000
0
Al particles (µg L-1)
1000
2000
3000
4000
5000
Fe particles (µg L-1)
Fig. 21. a) Amount of simulated Al in the form of Al(OH)3 (s) against the
amount of particulate (> 0.4 µm) aluminium for all samples with pH above
5.5 in the data set Krycklan. b) Amount of simulated Fe in the form of
ferrihydrite against the amount of particulate (> 0.4 µm) iron for all samples
with pH above 5.5 in the data set Krycklan.
to the field. Due to inherent differences in the techniques it is
always an advantage to use several methods simultaneously and to
compare the results (Sigg et al., 2006).
4.5. Geochemical equilibrium modelling ² is it possible to simulate
pH and metal speciation in lakes?
The pH simulations indicated no significant bias when using an
ADOM/DOC ratio of 1.65 considering all the modelled pH data.
The ratio (1.65) is close to the reported range of other authors
(Bryan et al., 2002: 1.12-1.98;; Tipping et al., 2002: 1.2-1.4;; Tipping
and Carter, 2011: 0.6-1.35). The calculated site density of 11.5 µM
mg-1 C is very close to the value of 10.2 µM mg-1 C proposed by
120
8
100
Ali simulted (µg L-1)
pH simulated
7
6
5
4
80
60
40
20
0
3
3
4
5
6
7
8
0
20
40
60
80
100
120
Ali analysed (µg L-1)
pH analysed
Fig. 22. Results of simulations at 75, 50, 25 and 0 % lime effect for the six
lakes with monitored termination of liming. Filled triangles: Lakes limed
directly at the lake surface for the previous 10 years. Open diamonds: Lakes
that were also limed upstream for the previous 10 years.
a) Comparison of simulated pH versus analysed pH.
b) Comparison of simulated Ali versus analysed Ali.
39
Carin Sjöstedt
TRITA LWR PHD 1065
(No limed lakes)
(No limed lakes)
0 to 10 %
10 to 40 %
40 to 50 %
50 to 70 %
0 to 10 %
10 to 40 %
40 to 50 %
50 to 70 %
Fig. 23 a) Percentages of limed lakes in the counties that have predicted pH
values below 5.6 for the no-lime scenario. b) Percentages of limed lakes in
the counties that have predicted Ali above 30 µg L-1 for the no-lime scenario.
+UXäka et al. (2003). The remaining random error in modelled pH
is due to the measurement errors from the 9 different
concentrations used as input parameters (Tipping et al., 1991). In
order to achieve higher precision when modelling pH other
methods can be used (Köhler et al., 2000). However, for this study
it was important with a more advanced model to be able to model
pH and metal speciation simultaneously.
Simulated Ali concentrations were generally in good agreement
with measured values, with some difference in the determination
methods, as discussed in section 4.3. The uncertainty is higher for
Ali than for pH, both in the fractionation and in the modelling, so
therefore it is better to use pH as the calibration endpoint when
calibrating the ADOM/DOC ratio. It was also important to use
the same value for the ADOM/DOC ratio when modelling pH
and Ali. If not, competition effects and changes in parameters e.g.
pH would have become erroneously predicted when some
parameters were changed, such as Ca and Mg.
The model worked well for molybdenum in the Dalsland lakes. On
the other hand, in the excessively limed lakes too strong
adsorption of arsenic to ferrihydrite was predicted. Nevertheless,
the general picture (strong arsenic adsorption to ferrihydrite) is
probably true due to the high correlation between arsenic and iron
detected in Swedish streams (Wällstedt et al., 2010).
Copper and lead were generally simulated well in the Tyresta lakes.
Lead was predicted to adsorb more strongly to ferrihydrite than
40
Iron and aluminium speciation, Implications for geochemical modelling
mg S/ m2
mg S/ m2
0-250
250-500
500-750
750-1000
1000-1250
1250-1500
0-250
250-500
500-750
750-1000
1000-1250
1250-1500
Fig. 24 a) Total sulphur deposition 1998 (SMHI, 2012) in Sweden. b) Total
sulphur deposition 2009 (SMHI, 2012).
copper, as expected from the constants derived from the batch
experiments in Paper V. Lead has previously been shown a strong
relationship with iron in the Tyresta lakes (Wällstedt et al., 2009).
The strong adsorption of lead, enhanced in presence of phosphate,
could explain the strong relationship between lead and iron found
in natural waters (Fuller et al., 1988;; Erel and Morgan, 1992).
However, complexation to DOM was important as well according
to the model and lead was also correlated to DOC in this study as
well as for other studies (Fuller et al., 1988).
4.6. The model as a tool for determining the extent of liming ² is it
possible to predict the future?
The model worked well as a tool to predict changes attributed to
change of lime dose in the six monitored previously limed lakes,
especially for the three lakes limed directly in the lake. The
modelling results of all limed lakes in Sweden indicated that it
would be possible to terminate liming in 30 % of the lakes. This
was especially true for the lakes in the inland areas of the northern
half of Sweden where almost all liming activities can be terminated.
In the coastal areas of northern Sweden liming can be terminated
in about 50 % of the lakes. The sulphur deposition has decreased
to a low level in northern Sweden (Fig. 24b) so it is not surprising
that the lakes are recovering from acidification. Also for the rest of
the lakes (in southern Sweden) it would be possible to reduce
liming. However, downscaling to a specific lake may not work for
the data in this study. More measurements of the water chemistry
have to be performed for specific lakes before termination of
liming. Changes of certain parameters, e.g. TOC may change in the
41
Carin Sjöstedt
TRITA LWR PHD 1065
future, which is not considered by the model. Also, the choice and
validity of the reference lakes are critical for the model
performance. However, the overall picture is probably true due to
the consistency of the results and of the sulphur deposition
patterns, and due also to the consistency of the results for the
unlimed reference lakes and of the six monitored previously limed
lakes. This would mean that the liming budget in Sweden of
208 million SEK can be significantly reduced.
4.7. Conclusions
Iron was characterised in two acidic, oligotrophic and humic lakes.
Iron(II) was occasionally present in high concentrations. Iron(III)
was found to be mainly complexed in monomeric complexes to
organic matter but also precipitated as an (hydr)oxide, likely
ferrihydrite. Iron(III) complexation was indicated to be strong,
and could therefore be a strong competitor with metals for
binding sites to DOM.
Aluminium solubility is likely controlled by a theoretical Al(OH)3
phase at higher pH in Swedish surface waters, but at lower pH
aluminium is undersaturated and instead controlled by
complexation to DOM. Therefore, for monitoring purposes total
aluminium needs to be determined in order to be able to model
Ali correctly.
Phosphate increased the adsorption of copper(II) and lead(II) to
ferrihydrite significantly, by forming ternary complexes with the
metals. The effect was higher for lead.
A geochemical equilibrium model was tested and evaluated for
freshwater all over Sweden. The model used one value for the
active DOM/DOC ratio of 1.65 and it required determined values
of total aluminium, iron, organic carbon, fluoride, sulphate, and
charge balance from major cations and anions in order to simulate
pH, Ali, Al(OH)3 and ferrihydrite simultaneously with reasonable
modelling errors. Trace metals such as Mo, As, Cu and Pb could
also be simulated.
The model could be used for predicting water chemistry after
termination of liming and to assess the amount of dissolved
inorganic fractions of different metals. The results suggested that
liming can be terminated in 30 % of the currently limed lakes in
Sweden and that liming can be decreased in many of the other
lakes.
Future studies should focus on comparing modelled Ali with
toxicity data of natural waters, to see if the model is a valuable tool
in predicting toxicity.
42
Iron and aluminium speciation, Implications for geochemical modelling
R EFERENCES
Alfaro-De la Torre, M.C., Beaulieu, P.-Y., and Tessier, A. (2000) In
situ measurement of trace metals in lakewater using the dialysis
and DGT techniques. Analytica Chimica Acta, 418, 53-68.
Andrén, C., and Rydin, E. (2009) Which aluminium fractionation
method will give true inorganic monomeric Al results in fresh
waters (not including colloidal Al)? Journal of Environmental
Monitoring, 11, 1639-1646.
Backes, C.A., and Tipping, E. (1987) An evaluation of the use of
cation-echange resin for the determination of organicallycomplexed Al in natural acid wates. International Journal of
Environmental Analytical Chemistry, 30, 135-143.
Barnes, R.B. (1975) The determination of specific forms of
aluminium in natural water. Chemical Geology, 15, 177-191.
Bidoglio, G., Grenthe, I., Qi, P., Robuch, P., and Omenetto, N.
(1991) Complexation of Eu and Tb with fulvic-acids as studied by
time-resolved laser-induced fluorescence. Talanta, 38 (9), 999-1008.
Björkvald, L., Buffam, I., Laudon, H., and Mörth, C.-M. (2008)
Hydrogeochemistry of Fe and Mn in small boreal streams: The
role of seasonality, landscape type and scale. Geochimica et
Cosmochimica Acta, 72, 2789-2804.
Bryan, S.E., Tipping, E., Hamilton-Taylor, J. (2002) Comparison of
measured and modeled copper binding by natural organic matter
in freshwaters. Comparative Biochemistry and Physiology Part C, 133, 3749.
Buffam, I., Laudon, H., Temrerud, J., and Mörth, C-M. (2007)
Landscape-scale variability of acidity and dissolved organic carbon
during spring flood in a boreal stream network. Journal of
Geophysical Research-Biosciences, 112 (G1), Article number: G01022
Cismasu, A.C., Michel, F.M., Tcaciuc, A.P., Tyliszczak, T., and Brown
Jr, G.E. (2011) Composition and structural aspects of naturally
occurring ferrihydrite. Comptes Rendus Geoscience, 343, 210-218.
Clarke, N., Danielsson, L.-G., and Sparén, A. (1992) The
determination of quickly reacting aluminium in natural waters by
kinetic discrimination in a flow system. International Journal of
Environmental Analytical Chemistry, 48 (2), 77-100.
Cory, N., Andrén, C.M., and Bishop, K. (2007) Modelling inorganic
aluminium with WHAM in environmental monitoring. Applied
Geochemistry, 22, 1196-1201.
Davison, W. (1993) Iron and manganese in lakes. Earth-Science Reviews,
34, 119-163.
di Toro, D.M., Allen, H.E., Bergman, H.L., Meyer, J.S., Paquin, P.R.,
and Santore, R.C. (2001) Biotic Ligand Model of the acute toxicity
of metals. 1. Technical basis. Environmental Toxicology and Chemistry,
20 (10), 2383-2396.
43
Carin Sjöstedt
TRITA LWR PHD 1065
Driscoll, C. T. (1984) A procedure for the fractionation of aqueous
aluminum in dilute acidic waters. International Journal of
Environmental Analytical Chemistry, 16 (4), 267²283.
Driscoll, C. T., Lehtinen, M.D., and Sullivan, T.J. (1994) Modeling the
acid-base chemistry of organic solutes in Adirondack, New York,
lakes. Water Resources Research, 30 (2), 297-306.
Driscoll, C.T., Driscoll, K.M., Roy, K.M., and Mitchell, M. (2003)
Chemical response of lakes in the Adirondack region of New York
to declines in acidic deposition. Environmental Science & Technology,
37, 2036-2042.
Dzombak, D.A., and Morel, F.M.M. (1990) Surface complexation
modelling: hydrous ferric oxide. Wiley-Interscience, New York,
393 pp.
Edkymish, A. (2009) Lead binding onto iron oxides (Ferrihydrite) as
influenced by aluminium and phosphorous. Masters Thesis in
Land and Water Resources Engineering, TRITA-LWR Masters
thesis 09:15
Erel, Y., and Morgan, J.J. (1992) The relationships between rockderived lead and iron in natural waters. Geochimica et Cosmochimica
Acta, 56, 4157-4167.
European Communities (2011) Common implementation strategy for
the Water Framework Directive (2000/60/EC). Guidance
Document no. 27.
Eusterhues, K., Wagner, F.E., Häusler, W., Hanzlik, M, Knicker, H.,
Totsche, K.U., Kögel-Knabner, I., and Schwertmann, U. (2008)
Characterization of ferrihydrite-soil organic matter coprecipitates
by X-ray diffraction and Mössbauer spectroscopy. Environmental
Science & Technology, 42, 7891-7897.
Evans, C.D., Monteith, D.T., and Cooper, D.M. (2005) Long-term
increases in surface water dissolved organic carbon: Observations,
possible causes and environmental impacts. Environmental Pollution,
137, 55-71.
Fölster, J. and Wilander, A. (2005) Försurningsbedömning i kalkade
vatten med kvoten Ca*/Mg*. Report 2005:3. Institutionen för
Miljöanalys, SLU (Swedish University of Agricultural Sciences),
SE-750 07 Uppsala, Sweden (in Swedish). 38 pp.
Fölster, J., Moldan, F. and Stadmark, J. (2011a) Målsjöundersökningen 2007-2008. Report 6412. June 2011. Swedish
Environmental Protection Agency, Stockholm, Sweden. (in
Swedish with English summary). 77 pp.
Fölster, J., Köhler, S., von Brömsen, C., Akselsson, C., and
Rönnback, P. (2011b) Korrigering av vattenkemi för
kalkningspåverkan: Val av referenser och beräkning av
osäkerheter. Report 2011:1 Department of Aquatic Sciences and
Assessment, SLU (Swedish University of Agricultural Sciences),
SE-750 07 Uppsala, Sweden (in Swedish) 69 pp.
Fuller, R.D., Simone, D.M., and Driscoll, C.T. (1988) Forest
clearcutting effects on trace metal concentrations: spatial patterns
44
Iron and aluminium speciation, Implications for geochemical modelling
in soil solutions and streams. Water, air and soil pollution, 40 (1-2),
185-195.
Gensemer, R.W., and Playle, R.C. (1999) The bioavailability and
toxicity of aluminium in aquatic environments. Critical Reviews in
Environmental Science & Technology, 29 (4), 315-450.
Gimpel, J., Zhang, H., Davison, W., and Edwards. A.C. (2003) In situ
trace metal speciation in lake surface waters using DGT, dialysis,
and filtration. Environmental Science & Technology, 37, 138-146.
Göransson E., Bringmark, E., Rapp, L., and Wilander, A. (2006)
Modeling the effect of liming on calcium concentration in Swedish
lakes. Environmental Monitoring and Assessment, 119 (1-3), 331-348.
Granier, C., BessagQHW % %RQG 7 '·$QJLROD $, Denier van der
Gon, H., Frost, G.J., Heil, A., Kaiser, J.W., Kinne, S., Klimont, Z.,
Kloster, S., Lamarque, J.-F., Liousse, C., Masui, T., Meleux, F.,
Mieville, A., Ohara, T., Raut, J.-C., Riahi, K., Schultz, M.G., Smith,
S.J., Thompson, A., van Ardenne, J., van der Werf, G.R., and van
Vuuren, D.P. (2011) Evolution of anthropogenic and biomass
burning emissions of air pollutants at global and regional scales
during the 1980-2010 period. Climate Change, 109, 163-190.
Gustafsson, J.P. (2001) Modeling the acid-base properties and metal
complexation of humic substances with the Stockholm Humic
Model. Journal of Colloid and Interface Science, 244 (1), 102-112.
Gustafsson, J.P., Berggren, D., Simonsson, M., Zysset, M., and
Mulder, J. (2001) Aluminium solubility mechanisms in moderately
acid Bs horizons of podzolized soils. European Journal of Soil Science,
52, 655-665.
Gustafsson, J.P., Persson I., Kleja D. B., and van Schaik J.W.J. (2007)
Binding of iron(III) to organic soils: EXAFS spectroscopy and
chemical equilibrium modeling. Environmental Science & Technology,
41, 1232-1237.
Gustafsson, J.P., Tiberg, C., Edkymish, A., and Kleja, D.B. (2011)
Modelling lead(II) sorption to ferrihydrite and soil organic matter.
Environmental Chemistry, 8 (5), 485-492.
Hedges, J.I., Eglinton, G., Hatcher, P.G., Kirchman, D.L., Arnosti,
C., Derenne, S., Evershed, R.P., Kögel-Knabner, I., de Leeuw,
J.W., Littke, R., Michaelis, W., and Rullkötter, J. (2000) The
molecularly-uncharacterized component of nonliving organic
matter in natural environments. Organic geochemistry, 31, 945-958.
Hiemstra, T., and van Riemsdijk, W.H. (1996) A surface structural
approach to ion adsorption: The Charge Distribution (CD) model.
Journal of Colloid and Interface Science, 179, 488-508.
Hiemstra, T., de Wit, J.C.M., and van Riemsdijk, W.H. (1989a)
Multisite proton adsorption modeling at the solid solution
interface of (hydr)oxides: A new approach. II. Application to
various important (hydr)oxides. Journal of Colloid and Interface Science,
133, 105-117.
Hiemstra, T., van Riemsdijk W.H., and Bolt, G.H. (1989b) Multisite
proton adsorption modeling at the solid/solution interface of
(hydr)oxides: A new approach. I. Model description and evaluation
45
Carin Sjöstedt
TRITA LWR PHD 1065
of intrinsic reaction constants. Journal of Colloid and Interface Science,
133, 91-104.
+UXäND - .|KOHU 6 /DXGRQ +, and Bishop, K. (2003) Is a
universal model of organic acidity possible: Comparison of the
acid/base properties of dissolved organic carbon in the boreal and
temperate zones. Environmental Science & Technology, 37, 1726-1730.
Hutchinson, N.J., and Sprague, J.B. (1986) Toxicity of trace metal
mixtures to American flagfish (Jordanella floridae) in soft, acidic
water and implications for cultural acidification. Canadian Journal of
Fisheries and Aquatic Sciences, 43, 647-655.
Karlsson T., and Persson P. (2010) Coordination chemistry and
hydrolysis of Fe(III) in a peat humic acid studied by X-ray
absorption spectroscopy. Geochimica et Cosmochimica Acta, 74 (1), 3040.
Karlsson T., Persson P., Skyllberg U., Mörth C.-M., and Giesler R.
(2008) Characterization of iron(III) in organic soils using extended
X-ray absorption fine structure spectroscopy. Environmental Science
& Technology, 42, 5449-5454.
Kelly S.D., Hesterberg D., and Ravel B. (2008) Analysis of soils and
minerals using X-ray absorption spectroscopy. In: Ulery, A.L. and
Drees, L.R, (Ed.) Methods of Soils Analysis. Part 5. Mineralogical
Methods. Soil Science of America Book series No. 5. Madison,
WI., pp. 387-463.
Khan, S. and Nugegoda, D. (2007) Sensitivity of juvenile freshwater
crayfish Cherax destructor (Decapoda: Parastacidae) to trace metals.
Ecotoxicology and Environmental Safety, 68 (3), 463-469.
Kinniburgh, D.G., van Riemsdijk, W.H., Koopal, L.K., Borkovec, M.,
Benedetti, M.F., and Avena, M.J. (1999) Ion binding to natural
organic matter: competition, heterogeneity, stoichiometry and
thermodynamic consistency. Colloids and Surfaces A: Physicochemical
and Engineering Aspects, 151, 147-166.
Köhler, S., Laudon, H., Wilander, A., and Bishop, K. (2000)
Estimating organic acid dissociation in natural surface waters using
total alkalinity and TOC. Water Research, 34 (5), 1425-1434.
Langford, C.H., and Khan, T.R. (1975) Kinetics and equilibrium of
binding of Fe3+ by a fulvic acid: A study by stopped flow methods.
Canadian Journal of Chemistry, 53 (20), 2979-2984.
Lindsay;; W.L., and Walthall, P.M. (1996) The solubility of aluminum
in soils. In: Sposito, G. (Ed.) The Environmental Chemistry of
Aluminum (2nd Edition). New York, Lewis Publishers. pp. 333361.
Liu, Y.-T., and Hesterberg, D. (2011) Phosphate bonding on noncrystalline Al/Fe-hydroxide coprecipitates. Environmental Science &
Technology, 45, 6283-6289.
Liu, X., and Millero, F.J. (1999) The solubility of iron hydroxide in
sodium chloride solutions. Geochimica et Cosmochimica Acta, 63
(19/20), 3487-3497.
46
Iron and aluminium speciation, Implications for geochemical modelling
Löfgren, S., Cory, N., and Zetterberg, T. (2010) Aluminium
concentrations in Swedish forest streams and co-variations with
catchment characteristics. Environmental Monitoring and Assessment,
166(1-4), 609-624.
Lofts, S., Tipping, E., and Hamilton-Taylor, J. (2008) The chemical
speciation of Fe(III) in freshwaters. Aquatic Geochemistry, 14 (4),
337-358.
Lydersen, E., Löfgren, S., and Arnesen, R. T. (2002) Metals in
Scandinavian Surface Waters: Effects of acidification, liming, and
potential reacidification. Critical Reviews in Environmental Science &
Technology, 32 (2-3), 73-295.
Lyvén, B., Hassellöv, M., Turner, D.R., Haraldsson, C., and
Andersson, K. (2003) Competition between iron- and carbonbased colloidal carriers for trace metals in a freshwater assessed
using flow field-flow fractionation coupled to ICPMS. Geochimica et
Cosmochimica Acta, 67 (20), 3791-3802.
Michel, F.M., Ehm, L., Antao, S.M., Lee, P.L., Chupas, P.J., Liu, G.,
Strongin, D.R., Schoonen, M.A.A., Phillips, B.L., and Parise, J.B.
(2007) The structure of ferrihydrite, a nanocrystalline material.
Science, 316, 1726-1729.
Michel, F.M., Barrón, V., Torrent, J., Morales, M.P., Serna, C.J., Boily,
J.-F., Liu, Q., Ambrosini, A., Cismasu, A.C., and Brown Jr, G.E.
(2010) Ordered ferrimagnetic form of ferrihydrite reveals links
among structure, composition, and magnetism. Proceedings of the
National Academy of Sciences of the United States of America, 107 (7),
2787-2792.
Mikutta C., and Kretzschmar R. (2011) Spectroscopic evidence for
ternary complex formation between arsenate and ferric iron
complexes of humic substances. Environmental Science & Technology,
45, 9550-9557.
Miles C.J., and Brezonik P.L. (1981) Oxygen consumption in humiccolored waters by a photochemical ferrous-ferric catalytic cycle.
Environmental Science & Technology, 15(9), 1089-1095.
Morel, F.M.M. (1983) Principles of Aquatic Chemistry. John Wiley &
Sons, New York, 446 pp.
Mota, A.M., Rato, A., Brazia, D., and Simões Gonçalves M.L. (1996)
Competition of Al3+ in complexation of humic matter with Pb2+: A
comparative study with other ions. Environmental Science &
Technology, 30, 1970-1974.
Neal, C., Lofts, S., Evans, C.D., Reynolds, B., Tipping, E. and Neal,
M. (2008) Increasing iron concentrations in UK upland waters.
Aquatic Geochemistry, 14, 263-288.
Niyogi, S., and Wood, C.M. (2004) Biotic Ligand Model, a flexible
tool for developing site-specific water quality guidelines for metals.
Environmental Science & Technology, 38 (23), 6177-6192.
Oliver, B.G., Thurman, E.M and Malcolm, R.L. (1983) The
contribution of humic substances to the acidity of colored natural
waters. Geochimica et Cosmoshimica Acta, 47, 2031-2035.
47
Carin Sjöstedt
TRITA LWR PHD 1065
Palmer, D.A., and Wesolowski, D.J. (1992) Aluminium speciation and
equilibria in aqueous solution: II. The solubility of gibbsite in
acidic sodium chloride solutions from 30 to 70ºC. Geochimica et
Cosmochimica Acta, 56, 1093-1111.
Parkhurst, D.L. and Appelo, C.A.J. (1999) 8VHU·V JXLGH WR
PHREEQC (Version 2)- A computer program for speciation,
batch-reaction, one-dimensional transport, and inverse
geochemical calculations: U.S. Geological Survey Water-Resources
Investigations Report 99-4259, 310 pp.
Perret, D., Gaillard, J.-F., Dominik, J., and Atteia, O. (2000) The
diversity of natural hydrous iron oxides. Environmental Science &
Technology, 34, 3540-3546.
Pinheiro, J.P., Mota, A.M., and Benedetti, M.F. (2000) Effect of
aluminium competition on lead and cadmium binding to humic
acids at variable ionic strength. Environmental Science & Technology,
34, 5137-5143.
Pokrovsky, O.S., and Schott, J. (2002) Iron colloids/organic matter
associated transport of major and trace elements in small boreal
rivers and their estuaries (NW Russia). Chemical Geology, 190, 141179.
Ravel, B. (2001) ATOMS: crystallography for the X-ray absorption
spectroscopist. Journal of Synchrotron Radiation, 8, 314-316.
Ravel, B. and Newville, M. (2005) ATHENA ARTEMIS
HEPHAESTUS: data analysis for X-ray absorption spectroscopy
using IFEFFIT. Journal of Synchrotron Radiation, 12 (4), 537-541.
Rehr, J.J., and Albers, R.C. (2000) Theoretical approaches to x-ray
absorption fine structure. Reviews of Modern Physics, 72, 621-654.
Richards, J.G., Burnison, B.K., and Playle, R.C. (1999) Natural and
commercial dissolved organic matter protects against the
physiological effects of a combined cadmium and copper exposure
on rainbow trout (Oncorhynchus mykiss). Canadian Journal of Fisheries
and Aquatic Sciences, 56, 407-418.
Rose, J., Vilge, A., Olivie-Lauquet, G., Masion, A., Frechou, C., and
Bottero, J.-Y. (1998) Iron speciation in natural organic matter
colloids. Colloids and Surfaces A., 136, 11-19.
Schecher, W.D., and Driscoll, C.T. (1988) An evaluation of the
equilibrium calculations within acidification models: The effect of
uncertainty in measured chemical components. Water Resources
Research, 24 (4), 533-540.
Schecher, W.D. and Driscoll, C. T. (1995) ALCHEMI: A chemical
equilibrium model to assess the acid-base chemistry and speciation
of aluminium in dilute solutions. In: Loeppert, R., Schwab, A.P.,
and Goldberg, S. (Ed.), Chemical Equilibrium and Reaction Models, Soil
Science of America, Madison, Wisconsin, pp. 325-356.
Scheinost, A.C., Abend, S., Pandya, K.I., and Sparks, D.L. (2001)
Kinetic controls on Cu and Pb sorption by ferrihydrite.
Environmental Science & Technology, 35, 1090-1096.
48
Iron and aluminium speciation, Implications for geochemical modelling
Schwertmann, U., Wagner, F., and Knicker, H. (2005) Ferrihydritehumic associations: magnetic hyperfine interactions. Soil Science
Society of America Journal, 69, 1009-1015.
Sigg, L., Black, F., Buffle, J., Cao, J., Cleven, R., Davison, W.,
Galceran, J., Gunkel, P., Kalis, E., Kistler, D., Martin, M., Noël, S.,
Nur, Y., Odzak, N., Puy, J., van Riemsdijk, W., Temminghoff, E.,
Tercier-Waeber, M.-L., Toepperwien, S., Town, R.M., Unsworth,
E., Warnken, K.W., Weng, L., Xue, H., and Zhang, H. (2006)
Comparison of analytical techniques for dynamic trace metal
speciation in natural freshwaters. Environmental Science & Technology,
40, 1934-1941.
Smith, R.M., Martell, A.E., and Motekaitis R.J. (2003) NIST critically
selected stability constants of metal complexes database. Version
7.0 NIST Standard Reference Database 46. Gaithersburg, National
Institute of Standards and Technology, US Department of
Commerce
Sobek, S., Algesten, G., Bergström, A.-K., Jansson, M., and Tranvik
L.J. (2003) The catchment and climate regulation of pCO2 in
boreal lakes. Global Change Biology, 9 (4), 630-641.
Stachowicz, M. (2007) Solubility of arsenic in multi-component
systems. From the microscopic to the macroscopic scale. Doctoral
thesis at Wageningen University, Wageningen, The Netherlands,
210 pp.
Swedish Environmental Protection Agency (EPA). (2011)
Naturvårdsverkets årsredovisning 2011, 129 pp.
Swedlund, P.J., Webster, J.G., and Miskelly, G.M. (2003) The effect of
SO4 on the ferrihydrite adsorption of Co, Pb and Cd: ternary
complexes and site heterogeneity. Applied Geochemistry, 18, 16711689.
Sunda, W.G., and Guillard, R.R.L. (1976) Relationship between cupric
ion activity and toxicity of copper to phytoplankton. Journal of
Marine Research, 34 (4), 511-529.
Susetyo, W., Dobbs, J.C., Carreira, L.A., Azarraga, L.V., and Grimm,
D.M. (1990) Development of a statistical model for metal-humic
interactions. Analytical Chemistry, 62, 1215-1221.
Teien, H.-C., Salbu, B., Kroglund, F., Heier, L.S., and Rosseland, B.O.
(2007) The influence on colloidal material on aluminium speciation
and estimated acid neutralizing capacity (ANC). Applied
Geochemistry, 22, 1202-1208.
Temnerud, J., Seibert, J., Jansson, M., and Bishop, K. (2007) Spatial
variation in discharge and concentrations of organic carbon in a
catchment network of boreal streams in northern Sweden. Journal
of Hydrology, 342, 72-87.
Tipping, E. (1981) The adsorption of aquatic humic substances by
iron oxides. Geochimica et Cosmochimica Acta, 45, 191-199.
Tipping, E. (1994) WHAM-a chemical equilibrium model and
computer code for waters, sediments, and soils incorporating a
discrete site/electrostatic model of ion-binding by humic
substances. Computers & Geosciences, 20 (6), 973-1023.
49
Carin Sjöstedt
TRITA LWR PHD 1065
Tipping, E. (1998) Humic ion-binding Model VI: an improved
description of the interactions of protons and metal ions with
humic substances. Aquatic Geochemistry, 4 (1), 3-48.
Tipping, E. (2005) Modelling Al competition for heavy metal binding
by dissolved organic matter in soil and surface waters of acid and
neutral pH. Geoderma, 127, 293-304.
Tipping, E., and Carter, H.T. (2011) Aluminium speciation in streams
and lakes of the UK Acid Waters Monitoring Network, modelled
with WHAM. Science of the Total Environment, 409, 1550-1558.
Tipping, E., and Hurley, M.A. (1992). A unifying model of cation
binding by humic substances. Geochimica et Cosmochimica Acta, 56
(10), 3627-3641.
Tipping, E., Woof, C., and Hurley, M.A. (1991) Humic substances in
acid surface waters;; Modelling aluminum binding, contribution to
ionic charge-balance and control of pH. Water Research, 25, 425435.
Tipping, E., Rey-Castro, C., Bryan, S.E., and Hamilton-Taylor, J.
(2002) Al(III) and Fe(III) binding by humic substances in
freshwaters, and implications for trace metal speciation. Geochimica
et Cosmochimica Acta, 66, 3211-3224.
Trivedi, P., Dyer, J.A., and Sparks, D.L. (2003) Lead sorption onto
ferrihydrite. 1. A macroscopic and spectroscopic assessment.
Environmental Science & Technology, 37, 908-914.
Unsworth, E., Warnken, K., Zhang, H., Davison, W., Black, F.,
Buffle, J., Cao, J., Cleven, R., Galceran, J., Gunkel, P., Kalis, E.,
Kistler, D., van Leeuwen, H., Martin, M., Noel, S., Nur, Y., Odzak,
N., Puy, J., van Riemsdijk, W., Sigg, L., Temminghoff, E., TercierWaeber, M-L., Toepperwien, S., Town, R., Weng, L., and Xue, H.
(2006) Model predictions of metal speciation in freshwaters
compared to measurements by in situ techniques. Environmental
Science & Technology, 40, 1942-1949.
US-Environmental Protection Agency (EPA). (2007) Aquatic life
ambient freshwater quality criteria-Copper 2007 Revision. Federal
Register Vol 72, No 35, 7983-7985.
van Dijk, H. (1971) Cation binding of humic acids. Geoderma, 5 (1),
53-67.
van Leeuwen, H.P., Town, R.M., Buffle, J., Cleven, R.F.M.J.,
Davison, W., Puy, J., van Riemsdijk, W.H., and Sigg, L. (2005)
Dynamic speciation analysis and bioavailability of metals in aquatic
systems. Environmental Science & Technology, 39 (22), 8545-8556.
van Schaik, J. W.J., Persson, I., Berggren Kleja, D., and Gustafssson,
J.P. EXAFS study on the reactions between iron and fulvic acid in
acid aqueous solutions. (2008) Environmental Science & Technology,
42, 2367-2373.
van Schaik, J., Kleja, D.B., and Gustafsson, J.P. (2010) Acid-base and
copper-binding properties of three organic matter fractions
isolated from a forest floor soil solutions. Geochimica et Cosmochimica
Acta, 74, 1391-1406.
50
Iron and aluminium speciation, Implications for geochemical modelling
Vasyukova, E.V., Pokrovsky, O.S., Viers, J., Oliva, P., Dupré, B.,
Martin, F., and Candaudap, F. (2010) Trace elements in organicand iron-rich surficial fluids of the boreal zone: Assessing colloidal
forms via dialysis and ultrafiltration. Geochimica et Cosmochimica Acta,
74, 449-468.
Vilgé-Ritter A., Rose J., Masion A., Bottero J.-Y., and Lainé J.-M.
(1999) Chemistry and structure of aggregates formed with Fe-salts
and natural organic matter. Colloids and Surfaces A., 147, 297-308.
Viollier E., Inglett P.W., Hunter K., Roychoudhury A.N., and van
Cappellen P. (2000) The ferrozine method revisited: Fe(II)/Fe(III)
determination in natural waters. Applied Geochemistry, 15, 785-70.
Wällstedt T., Edberg F., and Borg H. (2009) Long-term water
chemical trends in two Swedish lakes after termination of liming.
Science of the Total Environment, 407, 3554-3562.
Wällstedt, T., Björkvald, L., and Gustafsson, J.P. (2010) Increasing
concentrations of arsenic and vanadium in (southern) Swedish
streams. Applied Geochemistry, 25, 1162-1175.
Warby, R.A.F., Johnson, C.E., and Driscoll, C.T. (2008) Changes in
aluminum concentrations and speciation in lakes across the
northeastern U.S. following reductions in acidic deposition.
Environmental Science & Technology, 42, 8668-8674.
Wilkinson, K., Campbell, P.G.C., and Couture, P. (1990) Effect of
fluoride complexation on aluminium toxicity towards juvenile
Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic
Sciences, 47, 1446-1452.
Xie, L., and Giammar, D.E. (2007) Influence of phosphate on
adsorption and surface precipitation of lead on iron oxide surfaces.
In: Barnett, M.O. and Kent, D.O. (Ed.), Adsorption of Metals by
Geomedia II: Variables, mechanisms, and model applications,
Elsevier, Amsterdam, The Netherlands, pp. 349-373.
Other references
Chukalina, M. (2010). Wavelet2.ipf, a procedure for calculating the
Wavelet transform in IGOR Pro. Downloaded from website in
November 2011:
http://www.esrf.eu/UsersAndScience/Experiments/CRG/BM20
/Software/Wavelets/IGOR
Gustafsson J.P. (2012) Visual MINTEQ program.
http://www2.lwr.kth.se/english/OurSoftWare/Vminteq/index.ht
ml
ISELAW. (2007), http://info1.ma.slu.se/IKEU
MINEQL+ program (www.mineql.com)
SMHI (Swedish Meteorological and Hydrological Institute). (2012)
National survey RI DWPRVSKHULF FKHPLFDO GDWD IRU 6ZHGHQ·V
environmental monitoring, version Atmosfärskemi2011.1,
published by SMHI on commission of the Swedish EPA. Available
at the website www.smhi.se March 2012.
51